From 27319d45323f6f7690d756ac16387c46b69d7618 Mon Sep 17 00:00:00 2001 From: Jordan Cristiano Date: Sat, 12 Mar 2016 17:08:51 -0500 Subject: [PATCH] Added snappy-1.1.3 compression library --- external/snappy-1.1.3/AUTHORS | 1 + external/snappy-1.1.3/COPYING | 54 + external/snappy-1.1.3/INSTALL | 370 ++++ external/snappy-1.1.3/NEWS | 140 ++ external/snappy-1.1.3/README | 135 ++ .../snappy-1.1.3/include/snappy/snappy-c.h | 138 ++ .../include/snappy/snappy-stubs-public.h | 103 ++ external/snappy-1.1.3/include/snappy/snappy.h | 203 +++ external/snappy-1.1.3/snappy-c.cc | 90 + external/snappy-1.1.3/snappy-internal.h | 150 ++ external/snappy-1.1.3/snappy-sinksource.cc | 104 ++ external/snappy-1.1.3/snappy-sinksource.h | 182 ++ .../snappy-1.1.3/snappy-stubs-internal.cc | 42 + external/snappy-1.1.3/snappy-stubs-internal.h | 491 ++++++ external/snappy-1.1.3/snappy.cc | 1553 +++++++++++++++++ 15 files changed, 3756 insertions(+) create mode 100644 external/snappy-1.1.3/AUTHORS create mode 100644 external/snappy-1.1.3/COPYING create mode 100644 external/snappy-1.1.3/INSTALL create mode 100644 external/snappy-1.1.3/NEWS create mode 100644 external/snappy-1.1.3/README create mode 100644 external/snappy-1.1.3/include/snappy/snappy-c.h create mode 100644 external/snappy-1.1.3/include/snappy/snappy-stubs-public.h create mode 100644 external/snappy-1.1.3/include/snappy/snappy.h create mode 100644 external/snappy-1.1.3/snappy-c.cc create mode 100644 external/snappy-1.1.3/snappy-internal.h create mode 100644 external/snappy-1.1.3/snappy-sinksource.cc create mode 100644 external/snappy-1.1.3/snappy-sinksource.h create mode 100644 external/snappy-1.1.3/snappy-stubs-internal.cc create mode 100644 external/snappy-1.1.3/snappy-stubs-internal.h create mode 100644 external/snappy-1.1.3/snappy.cc diff --git a/external/snappy-1.1.3/AUTHORS b/external/snappy-1.1.3/AUTHORS new file mode 100644 index 0000000..4858b37 --- /dev/null +++ b/external/snappy-1.1.3/AUTHORS @@ -0,0 +1 @@ +opensource@google.com diff --git a/external/snappy-1.1.3/COPYING b/external/snappy-1.1.3/COPYING new file mode 100644 index 0000000..bd0e597 --- /dev/null +++ b/external/snappy-1.1.3/COPYING @@ -0,0 +1,54 @@ +Copyright 2011, Google Inc. +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + * Redistributions of source code must retain the above copyright +notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above +copyright notice, this list of conditions and the following disclaimer +in the documentation and/or other materials provided with the +distribution. + * Neither the name of Google Inc. nor the names of its +contributors may be used to endorse or promote products derived from +this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +=== + +Some of the benchmark data in testdata/ is licensed differently: + + - fireworks.jpeg is Copyright 2013 Steinar H. Gunderson, and + is licensed under the Creative Commons Attribution 3.0 license + (CC-BY-3.0). See https://creativecommons.org/licenses/by/3.0/ + for more information. + + - kppkn.gtb is taken from the Gaviota chess tablebase set, and + is licensed under the MIT License. See + https://sites.google.com/site/gaviotachessengine/Home/endgame-tablebases-1 + for more information. + + - paper-100k.pdf is an excerpt (bytes 92160 to 194560) from the paper + “Combinatorial Modeling of Chromatin Features Quantitatively Predicts DNA + Replication Timing in _Drosophila_” by Federico Comoglio and Renato Paro, + which is licensed under the CC-BY license. See + http://www.ploscompbiol.org/static/license for more ifnormation. + + - alice29.txt, asyoulik.txt, plrabn12.txt and lcet10.txt are from Project + Gutenberg. The first three have expired copyrights and are in the public + domain; the latter does not have expired copyright, but is still in the + public domain according to the license information + (http://www.gutenberg.org/ebooks/53). diff --git a/external/snappy-1.1.3/INSTALL b/external/snappy-1.1.3/INSTALL new file mode 100644 index 0000000..2099840 --- /dev/null +++ b/external/snappy-1.1.3/INSTALL @@ -0,0 +1,370 @@ +Installation Instructions +************************* + +Copyright (C) 1994-1996, 1999-2002, 2004-2013 Free Software Foundation, +Inc. + + Copying and distribution of this file, with or without modification, +are permitted in any medium without royalty provided the copyright +notice and this notice are preserved. This file is offered as-is, +without warranty of any kind. + +Basic Installation +================== + + Briefly, the shell command `./configure && make && make install' +should configure, build, and install this package. The following +more-detailed instructions are generic; see the `README' file for +instructions specific to this package. Some packages provide this +`INSTALL' file but do not implement all of the features documented +below. The lack of an optional feature in a given package is not +necessarily a bug. More recommendations for GNU packages can be found +in *note Makefile Conventions: (standards)Makefile Conventions. + + The `configure' shell script attempts to guess correct values for +various system-dependent variables used during compilation. It uses +those values to create a `Makefile' in each directory of the package. +It may also create one or more `.h' files containing system-dependent +definitions. Finally, it creates a shell script `config.status' that +you can run in the future to recreate the current configuration, and a +file `config.log' containing compiler output (useful mainly for +debugging `configure'). + + It can also use an optional file (typically called `config.cache' +and enabled with `--cache-file=config.cache' or simply `-C') that saves +the results of its tests to speed up reconfiguring. Caching is +disabled by default to prevent problems with accidental use of stale +cache files. + + If you need to do unusual things to compile the package, please try +to figure out how `configure' could check whether to do them, and mail +diffs or instructions to the address given in the `README' so they can +be considered for the next release. If you are using the cache, and at +some point `config.cache' contains results you don't want to keep, you +may remove or edit it. + + The file `configure.ac' (or `configure.in') is used to create +`configure' by a program called `autoconf'. You need `configure.ac' if +you want to change it or regenerate `configure' using a newer version +of `autoconf'. + + The simplest way to compile this package is: + + 1. `cd' to the directory containing the package's source code and type + `./configure' to configure the package for your system. + + Running `configure' might take a while. While running, it prints + some messages telling which features it is checking for. + + 2. Type `make' to compile the package. + + 3. Optionally, type `make check' to run any self-tests that come with + the package, generally using the just-built uninstalled binaries. + + 4. Type `make install' to install the programs and any data files and + documentation. When installing into a prefix owned by root, it is + recommended that the package be configured and built as a regular + user, and only the `make install' phase executed with root + privileges. + + 5. Optionally, type `make installcheck' to repeat any self-tests, but + this time using the binaries in their final installed location. + This target does not install anything. Running this target as a + regular user, particularly if the prior `make install' required + root privileges, verifies that the installation completed + correctly. + + 6. You can remove the program binaries and object files from the + source code directory by typing `make clean'. To also remove the + files that `configure' created (so you can compile the package for + a different kind of computer), type `make distclean'. There is + also a `make maintainer-clean' target, but that is intended mainly + for the package's developers. If you use it, you may have to get + all sorts of other programs in order to regenerate files that came + with the distribution. + + 7. Often, you can also type `make uninstall' to remove the installed + files again. In practice, not all packages have tested that + uninstallation works correctly, even though it is required by the + GNU Coding Standards. + + 8. Some packages, particularly those that use Automake, provide `make + distcheck', which can by used by developers to test that all other + targets like `make install' and `make uninstall' work correctly. + This target is generally not run by end users. + +Compilers and Options +===================== + + Some systems require unusual options for compilation or linking that +the `configure' script does not know about. Run `./configure --help' +for details on some of the pertinent environment variables. + + You can give `configure' initial values for configuration parameters +by setting variables in the command line or in the environment. Here +is an example: + + ./configure CC=c99 CFLAGS=-g LIBS=-lposix + + *Note Defining Variables::, for more details. + +Compiling For Multiple Architectures +==================================== + + You can compile the package for more than one kind of computer at the +same time, by placing the object files for each architecture in their +own directory. To do this, you can use GNU `make'. `cd' to the +directory where you want the object files and executables to go and run +the `configure' script. `configure' automatically checks for the +source code in the directory that `configure' is in and in `..'. This +is known as a "VPATH" build. + + With a non-GNU `make', it is safer to compile the package for one +architecture at a time in the source code directory. After you have +installed the package for one architecture, use `make distclean' before +reconfiguring for another architecture. + + On MacOS X 10.5 and later systems, you can create libraries and +executables that work on multiple system types--known as "fat" or +"universal" binaries--by specifying multiple `-arch' options to the +compiler but only a single `-arch' option to the preprocessor. Like +this: + + ./configure CC="gcc -arch i386 -arch x86_64 -arch ppc -arch ppc64" \ + CXX="g++ -arch i386 -arch x86_64 -arch ppc -arch ppc64" \ + CPP="gcc -E" CXXCPP="g++ -E" + + This is not guaranteed to produce working output in all cases, you +may have to build one architecture at a time and combine the results +using the `lipo' tool if you have problems. + +Installation Names +================== + + By default, `make install' installs the package's commands under +`/usr/local/bin', include files under `/usr/local/include', etc. You +can specify an installation prefix other than `/usr/local' by giving +`configure' the option `--prefix=PREFIX', where PREFIX must be an +absolute file name. + + You can specify separate installation prefixes for +architecture-specific files and architecture-independent files. If you +pass the option `--exec-prefix=PREFIX' to `configure', the package uses +PREFIX as the prefix for installing programs and libraries. +Documentation and other data files still use the regular prefix. + + In addition, if you use an unusual directory layout you can give +options like `--bindir=DIR' to specify different values for particular +kinds of files. Run `configure --help' for a list of the directories +you can set and what kinds of files go in them. In general, the +default for these options is expressed in terms of `${prefix}', so that +specifying just `--prefix' will affect all of the other directory +specifications that were not explicitly provided. + + The most portable way to affect installation locations is to pass the +correct locations to `configure'; however, many packages provide one or +both of the following shortcuts of passing variable assignments to the +`make install' command line to change installation locations without +having to reconfigure or recompile. + + The first method involves providing an override variable for each +affected directory. For example, `make install +prefix=/alternate/directory' will choose an alternate location for all +directory configuration variables that were expressed in terms of +`${prefix}'. Any directories that were specified during `configure', +but not in terms of `${prefix}', must each be overridden at install +time for the entire installation to be relocated. The approach of +makefile variable overrides for each directory variable is required by +the GNU Coding Standards, and ideally causes no recompilation. +However, some platforms have known limitations with the semantics of +shared libraries that end up requiring recompilation when using this +method, particularly noticeable in packages that use GNU Libtool. + + The second method involves providing the `DESTDIR' variable. For +example, `make install DESTDIR=/alternate/directory' will prepend +`/alternate/directory' before all installation names. The approach of +`DESTDIR' overrides is not required by the GNU Coding Standards, and +does not work on platforms that have drive letters. On the other hand, +it does better at avoiding recompilation issues, and works well even +when some directory options were not specified in terms of `${prefix}' +at `configure' time. + +Optional Features +================= + + If the package supports it, you can cause programs to be installed +with an extra prefix or suffix on their names by giving `configure' the +option `--program-prefix=PREFIX' or `--program-suffix=SUFFIX'. + + Some packages pay attention to `--enable-FEATURE' options to +`configure', where FEATURE indicates an optional part of the package. +They may also pay attention to `--with-PACKAGE' options, where PACKAGE +is something like `gnu-as' or `x' (for the X Window System). The +`README' should mention any `--enable-' and `--with-' options that the +package recognizes. + + For packages that use the X Window System, `configure' can usually +find the X include and library files automatically, but if it doesn't, +you can use the `configure' options `--x-includes=DIR' and +`--x-libraries=DIR' to specify their locations. + + Some packages offer the ability to configure how verbose the +execution of `make' will be. For these packages, running `./configure +--enable-silent-rules' sets the default to minimal output, which can be +overridden with `make V=1'; while running `./configure +--disable-silent-rules' sets the default to verbose, which can be +overridden with `make V=0'. + +Particular systems +================== + + On HP-UX, the default C compiler is not ANSI C compatible. If GNU +CC is not installed, it is recommended to use the following options in +order to use an ANSI C compiler: + + ./configure CC="cc -Ae -D_XOPEN_SOURCE=500" + +and if that doesn't work, install pre-built binaries of GCC for HP-UX. + + HP-UX `make' updates targets which have the same time stamps as +their prerequisites, which makes it generally unusable when shipped +generated files such as `configure' are involved. Use GNU `make' +instead. + + On OSF/1 a.k.a. Tru64, some versions of the default C compiler cannot +parse its `' header file. The option `-nodtk' can be used as +a workaround. If GNU CC is not installed, it is therefore recommended +to try + + ./configure CC="cc" + +and if that doesn't work, try + + ./configure CC="cc -nodtk" + + On Solaris, don't put `/usr/ucb' early in your `PATH'. This +directory contains several dysfunctional programs; working variants of +these programs are available in `/usr/bin'. So, if you need `/usr/ucb' +in your `PATH', put it _after_ `/usr/bin'. + + On Haiku, software installed for all users goes in `/boot/common', +not `/usr/local'. It is recommended to use the following options: + + ./configure --prefix=/boot/common + +Specifying the System Type +========================== + + There may be some features `configure' cannot figure out +automatically, but needs to determine by the type of machine the package +will run on. Usually, assuming the package is built to be run on the +_same_ architectures, `configure' can figure that out, but if it prints +a message saying it cannot guess the machine type, give it the +`--build=TYPE' option. TYPE can either be a short name for the system +type, such as `sun4', or a canonical name which has the form: + + CPU-COMPANY-SYSTEM + +where SYSTEM can have one of these forms: + + OS + KERNEL-OS + + See the file `config.sub' for the possible values of each field. If +`config.sub' isn't included in this package, then this package doesn't +need to know the machine type. + + If you are _building_ compiler tools for cross-compiling, you should +use the option `--target=TYPE' to select the type of system they will +produce code for. + + If you want to _use_ a cross compiler, that generates code for a +platform different from the build platform, you should specify the +"host" platform (i.e., that on which the generated programs will +eventually be run) with `--host=TYPE'. + +Sharing Defaults +================ + + If you want to set default values for `configure' scripts to share, +you can create a site shell script called `config.site' that gives +default values for variables like `CC', `cache_file', and `prefix'. +`configure' looks for `PREFIX/share/config.site' if it exists, then +`PREFIX/etc/config.site' if it exists. Or, you can set the +`CONFIG_SITE' environment variable to the location of the site script. +A warning: not all `configure' scripts look for a site script. + +Defining Variables +================== + + Variables not defined in a site shell script can be set in the +environment passed to `configure'. However, some packages may run +configure again during the build, and the customized values of these +variables may be lost. In order to avoid this problem, you should set +them in the `configure' command line, using `VAR=value'. For example: + + ./configure CC=/usr/local2/bin/gcc + +causes the specified `gcc' to be used as the C compiler (unless it is +overridden in the site shell script). + +Unfortunately, this technique does not work for `CONFIG_SHELL' due to +an Autoconf limitation. Until the limitation is lifted, you can use +this workaround: + + CONFIG_SHELL=/bin/bash ./configure CONFIG_SHELL=/bin/bash + +`configure' Invocation +====================== + + `configure' recognizes the following options to control how it +operates. + +`--help' +`-h' + Print a summary of all of the options to `configure', and exit. + +`--help=short' +`--help=recursive' + Print a summary of the options unique to this package's + `configure', and exit. The `short' variant lists options used + only in the top level, while the `recursive' variant lists options + also present in any nested packages. + +`--version' +`-V' + Print the version of Autoconf used to generate the `configure' + script, and exit. + +`--cache-file=FILE' + Enable the cache: use and save the results of the tests in FILE, + traditionally `config.cache'. FILE defaults to `/dev/null' to + disable caching. + +`--config-cache' +`-C' + Alias for `--cache-file=config.cache'. + +`--quiet' +`--silent' +`-q' + Do not print messages saying which checks are being made. To + suppress all normal output, redirect it to `/dev/null' (any error + messages will still be shown). + +`--srcdir=DIR' + Look for the package's source code in directory DIR. Usually + `configure' can determine that directory automatically. + +`--prefix=DIR' + Use DIR as the installation prefix. *note Installation Names:: + for more details, including other options available for fine-tuning + the installation locations. + +`--no-create' +`-n' + Run the configure checks, but stop before creating any output + files. + +`configure' also accepts some other, not widely useful, options. Run +`configure --help' for more details. diff --git a/external/snappy-1.1.3/NEWS b/external/snappy-1.1.3/NEWS new file mode 100644 index 0000000..4eb7a1d --- /dev/null +++ b/external/snappy-1.1.3/NEWS @@ -0,0 +1,140 @@ +Snappy v1.1.3, July 6th 2015: + +This is the first release to be done from GitHub, which means that +some minor things like the ChangeLog format has changed (git log +format instead of svn log). + + * Add support for Uncompress() from a Source to a Sink. + + * Various minor changes to improve MSVC support; in particular, + the unit tests now compile and run under MSVC. + + +Snappy v1.1.2, February 28th 2014: + +This is a maintenance release with no changes to the actual library +source code. + + * Stop distributing benchmark data files that have unclear + or unsuitable licensing. + + * Add support for padding chunks in the framing format. + + +Snappy v1.1.1, October 15th 2013: + + * Add support for uncompressing to iovecs (scatter I/O). + The bulk of this patch was contributed by Mohit Aron. + + * Speed up decompression by ~2%; much more so (~13-20%) on + a few benchmarks on given compilers and CPUs. + + * Fix a few issues with MSVC compilation. + + * Support truncated test data in the benchmark. + + +Snappy v1.1.0, January 18th 2013: + + * Snappy now uses 64 kB block size instead of 32 kB. On average, + this means it compresses about 3% denser (more so for some + inputs), at the same or better speeds. + + * libsnappy no longer depends on iostream. + + * Some small performance improvements in compression on x86 + (0.5–1%). + + * Various portability fixes for ARM-based platforms, for MSVC, + and for GNU/Hurd. + + +Snappy v1.0.5, February 24th 2012: + + * More speed improvements. Exactly how big will depend on + the architecture: + + - 3–10% faster decompression for the base case (x86-64). + + - ARMv7 and higher can now use unaligned accesses, + and will see about 30% faster decompression and + 20–40% faster compression. + + - 32-bit platforms (ARM and 32-bit x86) will see 2–5% + faster compression. + + These are all cumulative (e.g., ARM gets all three speedups). + + * Fixed an issue where the unit test would crash on system + with less than 256 MB address space available, + e.g. some embedded platforms. + + * Added a framing format description, for use over e.g. HTTP, + or for a command-line compressor. We do not have any + implementations of this at the current point, but there seems + to be enough of a general interest in the topic. + Also make the format description slightly clearer. + + * Remove some compile-time warnings in -Wall + (mostly signed/unsigned comparisons), for easier embedding + into projects that use -Wall -Werror. + + +Snappy v1.0.4, September 15th 2011: + + * Speeded up the decompressor somewhat; typically about 2–8% + for Core i7, in 64-bit mode (comparable for Opteron). + Somewhat more for some tests, almost no gain for others. + + * Make Snappy compile on certain platforms it didn't before + (Solaris with SunPro C++, HP-UX, AIX). + + * Correct some minor errors in the format description. + + +Snappy v1.0.3, June 2nd 2011: + + * Speeded up the decompressor somewhat; about 3-6% for Core 2, + 6-13% for Core i7, and 5-12% for Opteron (all in 64-bit mode). + + * Added compressed format documentation. This text is new, + but an earlier version from Zeev Tarantov was used as reference. + + * Only link snappy_unittest against -lz and other autodetected + libraries, not libsnappy.so (which doesn't need any such dependency). + + * Fixed some display issues in the microbenchmarks, one of which would + frequently make the test crash on GNU/Hurd. + + +Snappy v1.0.2, April 29th 2011: + + * Relicense to a BSD-type license. + + * Added C bindings, contributed by Martin Gieseking. + + * More Win32 fixes, in particular for MSVC. + + * Replace geo.protodata with a newer version. + + * Fix timing inaccuracies in the unit test when comparing Snappy + to other algorithms. + + +Snappy v1.0.1, March 25th 2011: + +This is a maintenance release, mostly containing minor fixes. +There is no new functionality. The most important fixes include: + + * The COPYING file and all licensing headers now correctly state that + Snappy is licensed under the Apache 2.0 license. + + * snappy_unittest should now compile natively under Windows, + as well as on embedded systems with no mmap(). + + * Various autotools nits have been fixed. + + +Snappy v1.0, March 17th 2011: + + * Initial version. diff --git a/external/snappy-1.1.3/README b/external/snappy-1.1.3/README new file mode 100644 index 0000000..3bc8888 --- /dev/null +++ b/external/snappy-1.1.3/README @@ -0,0 +1,135 @@ +Snappy, a fast compressor/decompressor. + + +Introduction +============ + +Snappy is a compression/decompression library. It does not aim for maximum +compression, or compatibility with any other compression library; instead, +it aims for very high speeds and reasonable compression. For instance, +compared to the fastest mode of zlib, Snappy is an order of magnitude faster +for most inputs, but the resulting compressed files are anywhere from 20% to +100% bigger. (For more information, see "Performance", below.) + +Snappy has the following properties: + + * Fast: Compression speeds at 250 MB/sec and beyond, with no assembler code. + See "Performance" below. + * Stable: Over the last few years, Snappy has compressed and decompressed + petabytes of data in Google's production environment. The Snappy bitstream + format is stable and will not change between versions. + * Robust: The Snappy decompressor is designed not to crash in the face of + corrupted or malicious input. + * Free and open source software: Snappy is licensed under a BSD-type license. + For more information, see the included COPYING file. + +Snappy has previously been called "Zippy" in some Google presentations +and the like. + + +Performance +=========== + +Snappy is intended to be fast. On a single core of a Core i7 processor +in 64-bit mode, it compresses at about 250 MB/sec or more and decompresses at +about 500 MB/sec or more. (These numbers are for the slowest inputs in our +benchmark suite; others are much faster.) In our tests, Snappy usually +is faster than algorithms in the same class (e.g. LZO, LZF, FastLZ, QuickLZ, +etc.) while achieving comparable compression ratios. + +Typical compression ratios (based on the benchmark suite) are about 1.5-1.7x +for plain text, about 2-4x for HTML, and of course 1.0x for JPEGs, PNGs and +other already-compressed data. Similar numbers for zlib in its fastest mode +are 2.6-2.8x, 3-7x and 1.0x, respectively. More sophisticated algorithms are +capable of achieving yet higher compression rates, although usually at the +expense of speed. Of course, compression ratio will vary significantly with +the input. + +Although Snappy should be fairly portable, it is primarily optimized +for 64-bit x86-compatible processors, and may run slower in other environments. +In particular: + + - Snappy uses 64-bit operations in several places to process more data at + once than would otherwise be possible. + - Snappy assumes unaligned 32- and 64-bit loads and stores are cheap. + On some platforms, these must be emulated with single-byte loads + and stores, which is much slower. + - Snappy assumes little-endian throughout, and needs to byte-swap data in + several places if running on a big-endian platform. + +Experience has shown that even heavily tuned code can be improved. +Performance optimizations, whether for 64-bit x86 or other platforms, +are of course most welcome; see "Contact", below. + + +Usage +===== + +Note that Snappy, both the implementation and the main interface, +is written in C++. However, several third-party bindings to other languages +are available; see the Google Code page at http://code.google.com/p/snappy/ +for more information. Also, if you want to use Snappy from C code, you can +use the included C bindings in snappy-c.h. + +To use Snappy from your own C++ program, include the file "snappy.h" from +your calling file, and link against the compiled library. + +There are many ways to call Snappy, but the simplest possible is + + snappy::Compress(input.data(), input.size(), &output); + +and similarly + + snappy::Uncompress(input.data(), input.size(), &output); + +where "input" and "output" are both instances of std::string. + +There are other interfaces that are more flexible in various ways, including +support for custom (non-array) input sources. See the header file for more +information. + + +Tests and benchmarks +==================== + +When you compile Snappy, snappy_unittest is compiled in addition to the +library itself. You do not need it to use the compressor from your own library, +but it contains several useful components for Snappy development. + +First of all, it contains unit tests, verifying correctness on your machine in +various scenarios. If you want to change or optimize Snappy, please run the +tests to verify you have not broken anything. Note that if you have the +Google Test library installed, unit test behavior (especially failures) will be +significantly more user-friendly. You can find Google Test at + + http://code.google.com/p/googletest/ + +You probably also want the gflags library for handling of command-line flags; +you can find it at + + http://code.google.com/p/google-gflags/ + +In addition to the unit tests, snappy contains microbenchmarks used to +tune compression and decompression performance. These are automatically run +before the unit tests, but you can disable them using the flag +--run_microbenchmarks=false if you have gflags installed (otherwise you will +need to edit the source). + +Finally, snappy can benchmark Snappy against a few other compression libraries +(zlib, LZO, LZF, FastLZ and QuickLZ), if they were detected at configure time. +To benchmark using a given file, give the compression algorithm you want to test +Snappy against (e.g. --zlib) and then a list of one or more file names on the +command line. The testdata/ directory contains the files used by the +microbenchmark, which should provide a reasonably balanced starting point for +benchmarking. (Note that baddata[1-3].snappy are not intended as benchmarks; they +are used to verify correctness in the presence of corrupted data in the unit +test.) + + +Contact +======= + +Snappy is distributed through Google Code. For the latest version, a bug tracker, +and other information, see + + http://code.google.com/p/snappy/ diff --git a/external/snappy-1.1.3/include/snappy/snappy-c.h b/external/snappy-1.1.3/include/snappy/snappy-c.h new file mode 100644 index 0000000..32aa0c6 --- /dev/null +++ b/external/snappy-1.1.3/include/snappy/snappy-c.h @@ -0,0 +1,138 @@ +/* + * Copyright 2011 Martin Gieseking . + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are + * met: + * + * * Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following disclaimer + * in the documentation and/or other materials provided with the + * distribution. + * * Neither the name of Google Inc. nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * Plain C interface (a wrapper around the C++ implementation). + */ + +#ifndef THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_C_H_ +#define THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_C_H_ + +#ifdef __cplusplus +extern "C" { +#endif + +#include + +/* + * Return values; see the documentation for each function to know + * what each can return. + */ +typedef enum { + SNAPPY_OK = 0, + SNAPPY_INVALID_INPUT = 1, + SNAPPY_BUFFER_TOO_SMALL = 2 +} snappy_status; + +/* + * Takes the data stored in "input[0..input_length-1]" and stores + * it in the array pointed to by "compressed". + * + * signals the space available in "compressed". + * If it is not at least equal to "snappy_max_compressed_length(input_length)", + * SNAPPY_BUFFER_TOO_SMALL is returned. After successful compression, + * contains the true length of the compressed output, + * and SNAPPY_OK is returned. + * + * Example: + * size_t output_length = snappy_max_compressed_length(input_length); + * char* output = (char*)malloc(output_length); + * if (snappy_compress(input, input_length, output, &output_length) + * == SNAPPY_OK) { + * ... Process(output, output_length) ... + * } + * free(output); + */ +snappy_status snappy_compress(const char* input, + size_t input_length, + char* compressed, + size_t* compressed_length); + +/* + * Given data in "compressed[0..compressed_length-1]" generated by + * calling the snappy_compress routine, this routine stores + * the uncompressed data to + * uncompressed[0..uncompressed_length-1]. + * Returns failure (a value not equal to SNAPPY_OK) if the message + * is corrupted and could not be decrypted. + * + * signals the space available in "uncompressed". + * If it is not at least equal to the value returned by + * snappy_uncompressed_length for this stream, SNAPPY_BUFFER_TOO_SMALL + * is returned. After successful decompression, + * contains the true length of the decompressed output. + * + * Example: + * size_t output_length; + * if (snappy_uncompressed_length(input, input_length, &output_length) + * != SNAPPY_OK) { + * ... fail ... + * } + * char* output = (char*)malloc(output_length); + * if (snappy_uncompress(input, input_length, output, &output_length) + * == SNAPPY_OK) { + * ... Process(output, output_length) ... + * } + * free(output); + */ +snappy_status snappy_uncompress(const char* compressed, + size_t compressed_length, + char* uncompressed, + size_t* uncompressed_length); + +/* + * Returns the maximal size of the compressed representation of + * input data that is "source_length" bytes in length. + */ +size_t snappy_max_compressed_length(size_t source_length); + +/* + * REQUIRES: "compressed[]" was produced by snappy_compress() + * Returns SNAPPY_OK and stores the length of the uncompressed data in + * *result normally. Returns SNAPPY_INVALID_INPUT on parsing error. + * This operation takes O(1) time. + */ +snappy_status snappy_uncompressed_length(const char* compressed, + size_t compressed_length, + size_t* result); + +/* + * Check if the contents of "compressed[]" can be uncompressed successfully. + * Does not return the uncompressed data; if so, returns SNAPPY_OK, + * or if not, returns SNAPPY_INVALID_INPUT. + * Takes time proportional to compressed_length, but is usually at least a + * factor of four faster than actual decompression. + */ +snappy_status snappy_validate_compressed_buffer(const char* compressed, + size_t compressed_length); + +#ifdef __cplusplus +} // extern "C" +#endif + +#endif /* THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_C_H_ */ diff --git a/external/snappy-1.1.3/include/snappy/snappy-stubs-public.h b/external/snappy-1.1.3/include/snappy/snappy-stubs-public.h new file mode 100644 index 0000000..7cc6a55 --- /dev/null +++ b/external/snappy-1.1.3/include/snappy/snappy-stubs-public.h @@ -0,0 +1,103 @@ +// Copyright 2011 Google Inc. All Rights Reserved. +// Author: sesse@google.com (Steinar H. Gunderson) +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +// Various type stubs for the open-source version of Snappy. +// +// This file cannot include config.h, as it is included from snappy.h, +// which is a public header. Instead, snappy-stubs-public.h is generated by +// from snappy-stubs-public.h.in at configure time. + +#ifndef THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_ +#define THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_ + +#if 1 +#include +#endif + +#if 1 +#include +#endif + +#if defined(_MSC_VER) && (_MSC_VER <= 1900) +#include +typedef SSIZE_T ssize_t; +#endif + +#if 0 +#include +#endif + +#define SNAPPY_MAJOR 1 +#define SNAPPY_MINOR 1 +#define SNAPPY_PATCHLEVEL 3 +#define SNAPPY_VERSION \ + ((SNAPPY_MAJOR << 16) | (SNAPPY_MINOR << 8) | SNAPPY_PATCHLEVEL) + +#include + +namespace snappy { + +#if 1 +typedef int8_t int8; +typedef uint8_t uint8; +typedef int16_t int16; +typedef uint16_t uint16; +typedef int32_t int32; +typedef uint32_t uint32; +typedef int64_t int64; +typedef uint64_t uint64; +#else +typedef signed char int8; +typedef unsigned char uint8; +typedef short int16; +typedef unsigned short uint16; +typedef int int32; +typedef unsigned int uint32; +typedef long long int64; +typedef unsigned long long uint64; +#endif + +typedef std::string string; + +#define DISALLOW_COPY_AND_ASSIGN(TypeName) \ + TypeName(const TypeName&); \ + void operator=(const TypeName&) + +#if !0 +// Windows does not have an iovec type, yet the concept is universally useful. +// It is simple to define it ourselves, so we put it inside our own namespace. +struct iovec { + void* iov_base; + size_t iov_len; +}; +#endif + +} // namespace snappy + +#endif // THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_ diff --git a/external/snappy-1.1.3/include/snappy/snappy.h b/external/snappy-1.1.3/include/snappy/snappy.h new file mode 100644 index 0000000..ec3880f --- /dev/null +++ b/external/snappy-1.1.3/include/snappy/snappy.h @@ -0,0 +1,203 @@ +// Copyright 2005 and onwards Google Inc. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +// A light-weight compression algorithm. It is designed for speed of +// compression and decompression, rather than for the utmost in space +// savings. +// +// For getting better compression ratios when you are compressing data +// with long repeated sequences or compressing data that is similar to +// other data, while still compressing fast, you might look at first +// using BMDiff and then compressing the output of BMDiff with +// Snappy. + +#ifndef THIRD_PARTY_SNAPPY_SNAPPY_H__ +#define THIRD_PARTY_SNAPPY_SNAPPY_H__ + +#include +#include + +#include "snappy/snappy-stubs-public.h" + +namespace snappy { + class Source; + class Sink; + + // ------------------------------------------------------------------------ + // Generic compression/decompression routines. + // ------------------------------------------------------------------------ + + // Compress the bytes read from "*source" and append to "*sink". Return the + // number of bytes written. + size_t Compress(Source* source, Sink* sink); + + // Find the uncompressed length of the given stream, as given by the header. + // Note that the true length could deviate from this; the stream could e.g. + // be truncated. + // + // Also note that this leaves "*source" in a state that is unsuitable for + // further operations, such as RawUncompress(). You will need to rewind + // or recreate the source yourself before attempting any further calls. + bool GetUncompressedLength(Source* source, uint32* result); + + // ------------------------------------------------------------------------ + // Higher-level string based routines (should be sufficient for most users) + // ------------------------------------------------------------------------ + + // Sets "*output" to the compressed version of "input[0,input_length-1]". + // Original contents of *output are lost. + // + // REQUIRES: "input[]" is not an alias of "*output". + size_t Compress(const char* input, size_t input_length, string* output); + + // Decompresses "compressed[0,compressed_length-1]" to "*uncompressed". + // Original contents of "*uncompressed" are lost. + // + // REQUIRES: "compressed[]" is not an alias of "*uncompressed". + // + // returns false if the message is corrupted and could not be decompressed + bool Uncompress(const char* compressed, size_t compressed_length, + string* uncompressed); + + // Decompresses "compressed" to "*uncompressed". + // + // returns false if the message is corrupted and could not be decompressed + bool Uncompress(Source* compressed, Sink* uncompressed); + + // This routine uncompresses as much of the "compressed" as possible + // into sink. It returns the number of valid bytes added to sink + // (extra invalid bytes may have been added due to errors; the caller + // should ignore those). The emitted data typically has length + // GetUncompressedLength(), but may be shorter if an error is + // encountered. + size_t UncompressAsMuchAsPossible(Source* compressed, Sink* uncompressed); + + // ------------------------------------------------------------------------ + // Lower-level character array based routines. May be useful for + // efficiency reasons in certain circumstances. + // ------------------------------------------------------------------------ + + // REQUIRES: "compressed" must point to an area of memory that is at + // least "MaxCompressedLength(input_length)" bytes in length. + // + // Takes the data stored in "input[0..input_length]" and stores + // it in the array pointed to by "compressed". + // + // "*compressed_length" is set to the length of the compressed output. + // + // Example: + // char* output = new char[snappy::MaxCompressedLength(input_length)]; + // size_t output_length; + // RawCompress(input, input_length, output, &output_length); + // ... Process(output, output_length) ... + // delete [] output; + void RawCompress(const char* input, + size_t input_length, + char* compressed, + size_t* compressed_length); + + // Given data in "compressed[0..compressed_length-1]" generated by + // calling the Snappy::Compress routine, this routine + // stores the uncompressed data to + // uncompressed[0..GetUncompressedLength(compressed)-1] + // returns false if the message is corrupted and could not be decrypted + bool RawUncompress(const char* compressed, size_t compressed_length, + char* uncompressed); + + // Given data from the byte source 'compressed' generated by calling + // the Snappy::Compress routine, this routine stores the uncompressed + // data to + // uncompressed[0..GetUncompressedLength(compressed,compressed_length)-1] + // returns false if the message is corrupted and could not be decrypted + bool RawUncompress(Source* compressed, char* uncompressed); + + // Given data in "compressed[0..compressed_length-1]" generated by + // calling the Snappy::Compress routine, this routine + // stores the uncompressed data to the iovec "iov". The number of physical + // buffers in "iov" is given by iov_cnt and their cumulative size + // must be at least GetUncompressedLength(compressed). The individual buffers + // in "iov" must not overlap with each other. + // + // returns false if the message is corrupted and could not be decrypted + bool RawUncompressToIOVec(const char* compressed, size_t compressed_length, + const struct iovec* iov, size_t iov_cnt); + + // Given data from the byte source 'compressed' generated by calling + // the Snappy::Compress routine, this routine stores the uncompressed + // data to the iovec "iov". The number of physical + // buffers in "iov" is given by iov_cnt and their cumulative size + // must be at least GetUncompressedLength(compressed). The individual buffers + // in "iov" must not overlap with each other. + // + // returns false if the message is corrupted and could not be decrypted + bool RawUncompressToIOVec(Source* compressed, const struct iovec* iov, + size_t iov_cnt); + + // Returns the maximal size of the compressed representation of + // input data that is "source_bytes" bytes in length; + size_t MaxCompressedLength(size_t source_bytes); + + // REQUIRES: "compressed[]" was produced by RawCompress() or Compress() + // Returns true and stores the length of the uncompressed data in + // *result normally. Returns false on parsing error. + // This operation takes O(1) time. + bool GetUncompressedLength(const char* compressed, size_t compressed_length, + size_t* result); + + // Returns true iff the contents of "compressed[]" can be uncompressed + // successfully. Does not return the uncompressed data. Takes + // time proportional to compressed_length, but is usually at least + // a factor of four faster than actual decompression. + bool IsValidCompressedBuffer(const char* compressed, + size_t compressed_length); + + // Returns true iff the contents of "compressed" can be uncompressed + // successfully. Does not return the uncompressed data. Takes + // time proportional to *compressed length, but is usually at least + // a factor of four faster than actual decompression. + // On success, consumes all of *compressed. On failure, consumes an + // unspecified prefix of *compressed. + bool IsValidCompressed(Source* compressed); + + // The size of a compression block. Note that many parts of the compression + // code assumes that kBlockSize <= 65536; in particular, the hash table + // can only store 16-bit offsets, and EmitCopy() also assumes the offset + // is 65535 bytes or less. Note also that if you change this, it will + // affect the framing format (see framing_format.txt). + // + // Note that there might be older data around that is compressed with larger + // block sizes, so the decompression code should not rely on the + // non-existence of long backreferences. + static const int kBlockLog = 16; + static const size_t kBlockSize = 1 << kBlockLog; + + static const int kMaxHashTableBits = 14; + static const size_t kMaxHashTableSize = 1 << kMaxHashTableBits; +} // end namespace snappy + +#endif // THIRD_PARTY_SNAPPY_SNAPPY_H__ diff --git a/external/snappy-1.1.3/snappy-c.cc b/external/snappy-1.1.3/snappy-c.cc new file mode 100644 index 0000000..016f0d5 --- /dev/null +++ b/external/snappy-1.1.3/snappy-c.cc @@ -0,0 +1,90 @@ +// Copyright 2011 Martin Gieseking . +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include "snappy/snappy.h" +#include "snappy/snappy-c.h" + +extern "C" { + +snappy_status snappy_compress(const char* input, + size_t input_length, + char* compressed, + size_t *compressed_length) { + if (*compressed_length < snappy_max_compressed_length(input_length)) { + return SNAPPY_BUFFER_TOO_SMALL; + } + snappy::RawCompress(input, input_length, compressed, compressed_length); + return SNAPPY_OK; +} + +snappy_status snappy_uncompress(const char* compressed, + size_t compressed_length, + char* uncompressed, + size_t* uncompressed_length) { + size_t real_uncompressed_length; + if (!snappy::GetUncompressedLength(compressed, + compressed_length, + &real_uncompressed_length)) { + return SNAPPY_INVALID_INPUT; + } + if (*uncompressed_length < real_uncompressed_length) { + return SNAPPY_BUFFER_TOO_SMALL; + } + if (!snappy::RawUncompress(compressed, compressed_length, uncompressed)) { + return SNAPPY_INVALID_INPUT; + } + *uncompressed_length = real_uncompressed_length; + return SNAPPY_OK; +} + +size_t snappy_max_compressed_length(size_t source_length) { + return snappy::MaxCompressedLength(source_length); +} + +snappy_status snappy_uncompressed_length(const char *compressed, + size_t compressed_length, + size_t *result) { + if (snappy::GetUncompressedLength(compressed, + compressed_length, + result)) { + return SNAPPY_OK; + } else { + return SNAPPY_INVALID_INPUT; + } +} + +snappy_status snappy_validate_compressed_buffer(const char *compressed, + size_t compressed_length) { + if (snappy::IsValidCompressedBuffer(compressed, compressed_length)) { + return SNAPPY_OK; + } else { + return SNAPPY_INVALID_INPUT; + } +} + +} // extern "C" diff --git a/external/snappy-1.1.3/snappy-internal.h b/external/snappy-1.1.3/snappy-internal.h new file mode 100644 index 0000000..0653dc6 --- /dev/null +++ b/external/snappy-1.1.3/snappy-internal.h @@ -0,0 +1,150 @@ +// Copyright 2008 Google Inc. All Rights Reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +// Internals shared between the Snappy implementation and its unittest. + +#ifndef THIRD_PARTY_SNAPPY_SNAPPY_INTERNAL_H_ +#define THIRD_PARTY_SNAPPY_SNAPPY_INTERNAL_H_ + +#include "snappy-stubs-internal.h" + +namespace snappy { +namespace internal { + +class WorkingMemory { + public: + WorkingMemory() : large_table_(NULL) { } + ~WorkingMemory() { delete[] large_table_; } + + // Allocates and clears a hash table using memory in "*this", + // stores the number of buckets in "*table_size" and returns a pointer to + // the base of the hash table. + uint16* GetHashTable(size_t input_size, int* table_size); + + private: + uint16 small_table_[1<<10]; // 2KB + uint16* large_table_; // Allocated only when needed + + DISALLOW_COPY_AND_ASSIGN(WorkingMemory); +}; + +// Flat array compression that does not emit the "uncompressed length" +// prefix. Compresses "input" string to the "*op" buffer. +// +// REQUIRES: "input_length <= kBlockSize" +// REQUIRES: "op" points to an array of memory that is at least +// "MaxCompressedLength(input_length)" in size. +// REQUIRES: All elements in "table[0..table_size-1]" are initialized to zero. +// REQUIRES: "table_size" is a power of two +// +// Returns an "end" pointer into "op" buffer. +// "end - op" is the compressed size of "input". +char* CompressFragment(const char* input, + size_t input_length, + char* op, + uint16* table, + const int table_size); + +// Return the largest n such that +// +// s1[0,n-1] == s2[0,n-1] +// and n <= (s2_limit - s2). +// +// Does not read *s2_limit or beyond. +// Does not read *(s1 + (s2_limit - s2)) or beyond. +// Requires that s2_limit >= s2. +// +// Separate implementation for x86_64, for speed. Uses the fact that +// x86_64 is little endian. +#if defined(ARCH_K8) +static inline int FindMatchLength(const char* s1, + const char* s2, + const char* s2_limit) { + assert(s2_limit >= s2); + int matched = 0; + + // Find out how long the match is. We loop over the data 64 bits at a + // time until we find a 64-bit block that doesn't match; then we find + // the first non-matching bit and use that to calculate the total + // length of the match. + while (PREDICT_TRUE(s2 <= s2_limit - 8)) { + if (UNALIGNED_LOAD64(s2) == UNALIGNED_LOAD64(s1 + matched)) { + s2 += 8; + matched += 8; + } else { + // On current (mid-2008) Opteron models there is a 3% more + // efficient code sequence to find the first non-matching byte. + // However, what follows is ~10% better on Intel Core 2 and newer, + // and we expect AMD's bsf instruction to improve. + uint64 x = UNALIGNED_LOAD64(s2) ^ UNALIGNED_LOAD64(s1 + matched); + int matching_bits = Bits::FindLSBSetNonZero64(x); + matched += matching_bits >> 3; + return matched; + } + } + while (PREDICT_TRUE(s2 < s2_limit)) { + if (s1[matched] == *s2) { + ++s2; + ++matched; + } else { + return matched; + } + } + return matched; +} +#else +static inline int FindMatchLength(const char* s1, + const char* s2, + const char* s2_limit) { + // Implementation based on the x86-64 version, above. + assert(s2_limit >= s2); + int matched = 0; + + while (s2 <= s2_limit - 4 && + UNALIGNED_LOAD32(s2) == UNALIGNED_LOAD32(s1 + matched)) { + s2 += 4; + matched += 4; + } + if (LittleEndian::IsLittleEndian() && s2 <= s2_limit - 4) { + uint32 x = UNALIGNED_LOAD32(s2) ^ UNALIGNED_LOAD32(s1 + matched); + int matching_bits = Bits::FindLSBSetNonZero(x); + matched += matching_bits >> 3; + } else { + while ((s2 < s2_limit) && (s1[matched] == *s2)) { + ++s2; + ++matched; + } + } + return matched; +} +#endif + +} // end namespace internal +} // end namespace snappy + +#endif // THIRD_PARTY_SNAPPY_SNAPPY_INTERNAL_H_ diff --git a/external/snappy-1.1.3/snappy-sinksource.cc b/external/snappy-1.1.3/snappy-sinksource.cc new file mode 100644 index 0000000..369a132 --- /dev/null +++ b/external/snappy-1.1.3/snappy-sinksource.cc @@ -0,0 +1,104 @@ +// Copyright 2011 Google Inc. All Rights Reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include + +#include "snappy-sinksource.h" + +namespace snappy { + +Source::~Source() { } + +Sink::~Sink() { } + +char* Sink::GetAppendBuffer(size_t length, char* scratch) { + return scratch; +} + +char* Sink::GetAppendBufferVariable( + size_t min_size, size_t desired_size_hint, char* scratch, + size_t scratch_size, size_t* allocated_size) { + *allocated_size = scratch_size; + return scratch; +} + +void Sink::AppendAndTakeOwnership( + char* bytes, size_t n, + void (*deleter)(void*, const char*, size_t), + void *deleter_arg) { + Append(bytes, n); + (*deleter)(deleter_arg, bytes, n); +} + +ByteArraySource::~ByteArraySource() { } + +size_t ByteArraySource::Available() const { return left_; } + +const char* ByteArraySource::Peek(size_t* len) { + *len = left_; + return ptr_; +} + +void ByteArraySource::Skip(size_t n) { + left_ -= n; + ptr_ += n; +} + +UncheckedByteArraySink::~UncheckedByteArraySink() { } + +void UncheckedByteArraySink::Append(const char* data, size_t n) { + // Do no copying if the caller filled in the result of GetAppendBuffer() + if (data != dest_) { + memcpy(dest_, data, n); + } + dest_ += n; +} + +char* UncheckedByteArraySink::GetAppendBuffer(size_t len, char* scratch) { + return dest_; +} + +void UncheckedByteArraySink::AppendAndTakeOwnership( + char* data, size_t n, + void (*deleter)(void*, const char*, size_t), + void *deleter_arg) { + if (data != dest_) { + memcpy(dest_, data, n); + (*deleter)(deleter_arg, data, n); + } + dest_ += n; +} + +char* UncheckedByteArraySink::GetAppendBufferVariable( + size_t min_size, size_t desired_size_hint, char* scratch, + size_t scratch_size, size_t* allocated_size) { + *allocated_size = desired_size_hint; + return dest_; +} + +} // namespace snappy diff --git a/external/snappy-1.1.3/snappy-sinksource.h b/external/snappy-1.1.3/snappy-sinksource.h new file mode 100644 index 0000000..8afcdaa --- /dev/null +++ b/external/snappy-1.1.3/snappy-sinksource.h @@ -0,0 +1,182 @@ +// Copyright 2011 Google Inc. All Rights Reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#ifndef THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_ +#define THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_ + +#include + +namespace snappy { + +// A Sink is an interface that consumes a sequence of bytes. +class Sink { + public: + Sink() { } + virtual ~Sink(); + + // Append "bytes[0,n-1]" to this. + virtual void Append(const char* bytes, size_t n) = 0; + + // Returns a writable buffer of the specified length for appending. + // May return a pointer to the caller-owned scratch buffer which + // must have at least the indicated length. The returned buffer is + // only valid until the next operation on this Sink. + // + // After writing at most "length" bytes, call Append() with the + // pointer returned from this function and the number of bytes + // written. Many Append() implementations will avoid copying + // bytes if this function returned an internal buffer. + // + // If a non-scratch buffer is returned, the caller may only pass a + // prefix of it to Append(). That is, it is not correct to pass an + // interior pointer of the returned array to Append(). + // + // The default implementation always returns the scratch buffer. + virtual char* GetAppendBuffer(size_t length, char* scratch); + + // For higher performance, Sink implementations can provide custom + // AppendAndTakeOwnership() and GetAppendBufferVariable() methods. + // These methods can reduce the number of copies done during + // compression/decompression. + + // Append "bytes[0,n-1] to the sink. Takes ownership of "bytes" + // and calls the deleter function as (*deleter)(deleter_arg, bytes, n) + // to free the buffer. deleter function must be non NULL. + // + // The default implementation just calls Append and frees "bytes". + // Other implementations may avoid a copy while appending the buffer. + virtual void AppendAndTakeOwnership( + char* bytes, size_t n, void (*deleter)(void*, const char*, size_t), + void *deleter_arg); + + // Returns a writable buffer for appending and writes the buffer's capacity to + // *allocated_size. Guarantees *allocated_size >= min_size. + // May return a pointer to the caller-owned scratch buffer which must have + // scratch_size >= min_size. + // + // The returned buffer is only valid until the next operation + // on this ByteSink. + // + // After writing at most *allocated_size bytes, call Append() with the + // pointer returned from this function and the number of bytes written. + // Many Append() implementations will avoid copying bytes if this function + // returned an internal buffer. + // + // If the sink implementation allocates or reallocates an internal buffer, + // it should use the desired_size_hint if appropriate. If a caller cannot + // provide a reasonable guess at the desired capacity, it should set + // desired_size_hint = 0. + // + // If a non-scratch buffer is returned, the caller may only pass + // a prefix to it to Append(). That is, it is not correct to pass an + // interior pointer to Append(). + // + // The default implementation always returns the scratch buffer. + virtual char* GetAppendBufferVariable( + size_t min_size, size_t desired_size_hint, char* scratch, + size_t scratch_size, size_t* allocated_size); + + private: + // No copying + Sink(const Sink&); + void operator=(const Sink&); +}; + +// A Source is an interface that yields a sequence of bytes +class Source { + public: + Source() { } + virtual ~Source(); + + // Return the number of bytes left to read from the source + virtual size_t Available() const = 0; + + // Peek at the next flat region of the source. Does not reposition + // the source. The returned region is empty iff Available()==0. + // + // Returns a pointer to the beginning of the region and store its + // length in *len. + // + // The returned region is valid until the next call to Skip() or + // until this object is destroyed, whichever occurs first. + // + // The returned region may be larger than Available() (for example + // if this ByteSource is a view on a substring of a larger source). + // The caller is responsible for ensuring that it only reads the + // Available() bytes. + virtual const char* Peek(size_t* len) = 0; + + // Skip the next n bytes. Invalidates any buffer returned by + // a previous call to Peek(). + // REQUIRES: Available() >= n + virtual void Skip(size_t n) = 0; + + private: + // No copying + Source(const Source&); + void operator=(const Source&); +}; + +// A Source implementation that yields the contents of a flat array +class ByteArraySource : public Source { + public: + ByteArraySource(const char* p, size_t n) : ptr_(p), left_(n) { } + virtual ~ByteArraySource(); + virtual size_t Available() const; + virtual const char* Peek(size_t* len); + virtual void Skip(size_t n); + private: + const char* ptr_; + size_t left_; +}; + +// A Sink implementation that writes to a flat array without any bound checks. +class UncheckedByteArraySink : public Sink { + public: + explicit UncheckedByteArraySink(char* dest) : dest_(dest) { } + virtual ~UncheckedByteArraySink(); + virtual void Append(const char* data, size_t n); + virtual char* GetAppendBuffer(size_t len, char* scratch); + virtual char* GetAppendBufferVariable( + size_t min_size, size_t desired_size_hint, char* scratch, + size_t scratch_size, size_t* allocated_size); + virtual void AppendAndTakeOwnership( + char* bytes, size_t n, void (*deleter)(void*, const char*, size_t), + void *deleter_arg); + + // Return the current output pointer so that a caller can see how + // many bytes were produced. + // Note: this is not a Sink method. + char* CurrentDestination() const { return dest_; } + private: + char* dest_; +}; + +} // namespace snappy + +#endif // THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_ diff --git a/external/snappy-1.1.3/snappy-stubs-internal.cc b/external/snappy-1.1.3/snappy-stubs-internal.cc new file mode 100644 index 0000000..6ed3343 --- /dev/null +++ b/external/snappy-1.1.3/snappy-stubs-internal.cc @@ -0,0 +1,42 @@ +// Copyright 2011 Google Inc. All Rights Reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include +#include + +#include "snappy-stubs-internal.h" + +namespace snappy { + +void Varint::Append32(string* s, uint32 value) { + char buf[Varint::kMax32]; + const char* p = Varint::Encode32(buf, value); + s->append(buf, p - buf); +} + +} // namespace snappy diff --git a/external/snappy-1.1.3/snappy-stubs-internal.h b/external/snappy-1.1.3/snappy-stubs-internal.h new file mode 100644 index 0000000..d344b6b --- /dev/null +++ b/external/snappy-1.1.3/snappy-stubs-internal.h @@ -0,0 +1,491 @@ +// Copyright 2011 Google Inc. All Rights Reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +// Various stubs for the open-source version of Snappy. + +#ifndef THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_ +#define THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include + +#include +#include +#include + +#ifdef HAVE_SYS_MMAN_H +#include +#endif + +#include "snappy/snappy-stubs-public.h" + +#if defined(__x86_64__) + +// Enable 64-bit optimized versions of some routines. +#define ARCH_K8 1 + +#endif + +// Needed by OS X, among others. +#ifndef MAP_ANONYMOUS +#define MAP_ANONYMOUS MAP_ANON +#endif + +// Pull in std::min, std::ostream, and the likes. This is safe because this +// header file is never used from any public header files. +using namespace std; + +// The size of an array, if known at compile-time. +// Will give unexpected results if used on a pointer. +// We undefine it first, since some compilers already have a definition. +#ifdef ARRAYSIZE +#undef ARRAYSIZE +#endif +#define ARRAYSIZE(a) (sizeof(a) / sizeof(*(a))) + +// Static prediction hints. +#ifdef HAVE_BUILTIN_EXPECT +#define PREDICT_FALSE(x) (__builtin_expect(x, 0)) +#define PREDICT_TRUE(x) (__builtin_expect(!!(x), 1)) +#else +#define PREDICT_FALSE(x) x +#define PREDICT_TRUE(x) x +#endif + +// This is only used for recomputing the tag byte table used during +// decompression; for simplicity we just remove it from the open-source +// version (anyone who wants to regenerate it can just do the call +// themselves within main()). +#define DEFINE_bool(flag_name, default_value, description) \ + bool FLAGS_ ## flag_name = default_value +#define DECLARE_bool(flag_name) \ + extern bool FLAGS_ ## flag_name + +namespace snappy { + +static const uint32 kuint32max = static_cast(0xFFFFFFFF); +static const int64 kint64max = static_cast(0x7FFFFFFFFFFFFFFFLL); + +// Potentially unaligned loads and stores. + +// x86 and PowerPC can simply do these loads and stores native. + +#if defined(__i386__) || defined(__x86_64__) || defined(__powerpc__) + +#define UNALIGNED_LOAD16(_p) (*reinterpret_cast(_p)) +#define UNALIGNED_LOAD32(_p) (*reinterpret_cast(_p)) +#define UNALIGNED_LOAD64(_p) (*reinterpret_cast(_p)) + +#define UNALIGNED_STORE16(_p, _val) (*reinterpret_cast(_p) = (_val)) +#define UNALIGNED_STORE32(_p, _val) (*reinterpret_cast(_p) = (_val)) +#define UNALIGNED_STORE64(_p, _val) (*reinterpret_cast(_p) = (_val)) + +// ARMv7 and newer support native unaligned accesses, but only of 16-bit +// and 32-bit values (not 64-bit); older versions either raise a fatal signal, +// do an unaligned read and rotate the words around a bit, or do the reads very +// slowly (trip through kernel mode). There's no simple #define that says just +// “ARMv7 or higher”, so we have to filter away all ARMv5 and ARMv6 +// sub-architectures. +// +// This is a mess, but there's not much we can do about it. + +#elif defined(__arm__) && \ + !defined(__ARM_ARCH_4__) && \ + !defined(__ARM_ARCH_4T__) && \ + !defined(__ARM_ARCH_5__) && \ + !defined(__ARM_ARCH_5T__) && \ + !defined(__ARM_ARCH_5TE__) && \ + !defined(__ARM_ARCH_5TEJ__) && \ + !defined(__ARM_ARCH_6__) && \ + !defined(__ARM_ARCH_6J__) && \ + !defined(__ARM_ARCH_6K__) && \ + !defined(__ARM_ARCH_6Z__) && \ + !defined(__ARM_ARCH_6ZK__) && \ + !defined(__ARM_ARCH_6T2__) + +#define UNALIGNED_LOAD16(_p) (*reinterpret_cast(_p)) +#define UNALIGNED_LOAD32(_p) (*reinterpret_cast(_p)) + +#define UNALIGNED_STORE16(_p, _val) (*reinterpret_cast(_p) = (_val)) +#define UNALIGNED_STORE32(_p, _val) (*reinterpret_cast(_p) = (_val)) + +// TODO(user): NEON supports unaligned 64-bit loads and stores. +// See if that would be more efficient on platforms supporting it, +// at least for copies. + +inline uint64 UNALIGNED_LOAD64(const void *p) { + uint64 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline void UNALIGNED_STORE64(void *p, uint64 v) { + memcpy(p, &v, sizeof v); +} + +#else + +// These functions are provided for architectures that don't support +// unaligned loads and stores. + +inline uint16 UNALIGNED_LOAD16(const void *p) { + uint16 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline uint32 UNALIGNED_LOAD32(const void *p) { + uint32 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline uint64 UNALIGNED_LOAD64(const void *p) { + uint64 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline void UNALIGNED_STORE16(void *p, uint16 v) { + memcpy(p, &v, sizeof v); +} + +inline void UNALIGNED_STORE32(void *p, uint32 v) { + memcpy(p, &v, sizeof v); +} + +inline void UNALIGNED_STORE64(void *p, uint64 v) { + memcpy(p, &v, sizeof v); +} + +#endif + +// This can be more efficient than UNALIGNED_LOAD64 + UNALIGNED_STORE64 +// on some platforms, in particular ARM. +inline void UnalignedCopy64(const void *src, void *dst) { + if (sizeof(void *) == 8) { + UNALIGNED_STORE64(dst, UNALIGNED_LOAD64(src)); + } else { + const char *src_char = reinterpret_cast(src); + char *dst_char = reinterpret_cast(dst); + + UNALIGNED_STORE32(dst_char, UNALIGNED_LOAD32(src_char)); + UNALIGNED_STORE32(dst_char + 4, UNALIGNED_LOAD32(src_char + 4)); + } +} + +// The following guarantees declaration of the byte swap functions. +#ifdef WORDS_BIGENDIAN + +#ifdef HAVE_SYS_BYTEORDER_H +#include +#endif + +#ifdef HAVE_SYS_ENDIAN_H +#include +#endif + +#ifdef _MSC_VER +#include +#define bswap_16(x) _byteswap_ushort(x) +#define bswap_32(x) _byteswap_ulong(x) +#define bswap_64(x) _byteswap_uint64(x) + +#elif defined(__APPLE__) +// Mac OS X / Darwin features +#include +#define bswap_16(x) OSSwapInt16(x) +#define bswap_32(x) OSSwapInt32(x) +#define bswap_64(x) OSSwapInt64(x) + +#elif defined(HAVE_BYTESWAP_H) +#include + +#elif defined(bswap32) +// FreeBSD defines bswap{16,32,64} in (already #included). +#define bswap_16(x) bswap16(x) +#define bswap_32(x) bswap32(x) +#define bswap_64(x) bswap64(x) + +#elif defined(BSWAP_64) +// Solaris 10 defines BSWAP_{16,32,64} in (already #included). +#define bswap_16(x) BSWAP_16(x) +#define bswap_32(x) BSWAP_32(x) +#define bswap_64(x) BSWAP_64(x) + +#else + +inline uint16 bswap_16(uint16 x) { + return (x << 8) | (x >> 8); +} + +inline uint32 bswap_32(uint32 x) { + x = ((x & 0xff00ff00UL) >> 8) | ((x & 0x00ff00ffUL) << 8); + return (x >> 16) | (x << 16); +} + +inline uint64 bswap_64(uint64 x) { + x = ((x & 0xff00ff00ff00ff00ULL) >> 8) | ((x & 0x00ff00ff00ff00ffULL) << 8); + x = ((x & 0xffff0000ffff0000ULL) >> 16) | ((x & 0x0000ffff0000ffffULL) << 16); + return (x >> 32) | (x << 32); +} + +#endif + +#endif // WORDS_BIGENDIAN + +// Convert to little-endian storage, opposite of network format. +// Convert x from host to little endian: x = LittleEndian.FromHost(x); +// convert x from little endian to host: x = LittleEndian.ToHost(x); +// +// Store values into unaligned memory converting to little endian order: +// LittleEndian.Store16(p, x); +// +// Load unaligned values stored in little endian converting to host order: +// x = LittleEndian.Load16(p); +class LittleEndian { + public: + // Conversion functions. +#ifdef WORDS_BIGENDIAN + + static uint16 FromHost16(uint16 x) { return bswap_16(x); } + static uint16 ToHost16(uint16 x) { return bswap_16(x); } + + static uint32 FromHost32(uint32 x) { return bswap_32(x); } + static uint32 ToHost32(uint32 x) { return bswap_32(x); } + + static bool IsLittleEndian() { return false; } + +#else // !defined(WORDS_BIGENDIAN) + + static uint16 FromHost16(uint16 x) { return x; } + static uint16 ToHost16(uint16 x) { return x; } + + static uint32 FromHost32(uint32 x) { return x; } + static uint32 ToHost32(uint32 x) { return x; } + + static bool IsLittleEndian() { return true; } + +#endif // !defined(WORDS_BIGENDIAN) + + // Functions to do unaligned loads and stores in little-endian order. + static uint16 Load16(const void *p) { + return ToHost16(UNALIGNED_LOAD16(p)); + } + + static void Store16(void *p, uint16 v) { + UNALIGNED_STORE16(p, FromHost16(v)); + } + + static uint32 Load32(const void *p) { + return ToHost32(UNALIGNED_LOAD32(p)); + } + + static void Store32(void *p, uint32 v) { + UNALIGNED_STORE32(p, FromHost32(v)); + } +}; + +// Some bit-manipulation functions. +class Bits { + public: + // Return floor(log2(n)) for positive integer n. Returns -1 iff n == 0. + static int Log2Floor(uint32 n); + + // Return the first set least / most significant bit, 0-indexed. Returns an + // undefined value if n == 0. FindLSBSetNonZero() is similar to ffs() except + // that it's 0-indexed. + static int FindLSBSetNonZero(uint32 n); + static int FindLSBSetNonZero64(uint64 n); + + private: + DISALLOW_COPY_AND_ASSIGN(Bits); +}; + +#ifdef HAVE_BUILTIN_CTZ + +inline int Bits::Log2Floor(uint32 n) { + return n == 0 ? -1 : 31 ^ __builtin_clz(n); +} + +inline int Bits::FindLSBSetNonZero(uint32 n) { + return __builtin_ctz(n); +} + +inline int Bits::FindLSBSetNonZero64(uint64 n) { + return __builtin_ctzll(n); +} + +#else // Portable versions. + +inline int Bits::Log2Floor(uint32 n) { + if (n == 0) + return -1; + int log = 0; + uint32 value = n; + for (int i = 4; i >= 0; --i) { + int shift = (1 << i); + uint32 x = value >> shift; + if (x != 0) { + value = x; + log += shift; + } + } + assert(value == 1); + return log; +} + +inline int Bits::FindLSBSetNonZero(uint32 n) { + int rc = 31; + for (int i = 4, shift = 1 << 4; i >= 0; --i) { + const uint32 x = n << shift; + if (x != 0) { + n = x; + rc -= shift; + } + shift >>= 1; + } + return rc; +} + +// FindLSBSetNonZero64() is defined in terms of FindLSBSetNonZero(). +inline int Bits::FindLSBSetNonZero64(uint64 n) { + const uint32 bottombits = static_cast(n); + if (bottombits == 0) { + // Bottom bits are zero, so scan in top bits + return 32 + FindLSBSetNonZero(static_cast(n >> 32)); + } else { + return FindLSBSetNonZero(bottombits); + } +} + +#endif // End portable versions. + +// Variable-length integer encoding. +class Varint { + public: + // Maximum lengths of varint encoding of uint32. + static const int kMax32 = 5; + + // Attempts to parse a varint32 from a prefix of the bytes in [ptr,limit-1]. + // Never reads a character at or beyond limit. If a valid/terminated varint32 + // was found in the range, stores it in *OUTPUT and returns a pointer just + // past the last byte of the varint32. Else returns NULL. On success, + // "result <= limit". + static const char* Parse32WithLimit(const char* ptr, const char* limit, + uint32* OUTPUT); + + // REQUIRES "ptr" points to a buffer of length sufficient to hold "v". + // EFFECTS Encodes "v" into "ptr" and returns a pointer to the + // byte just past the last encoded byte. + static char* Encode32(char* ptr, uint32 v); + + // EFFECTS Appends the varint representation of "value" to "*s". + static void Append32(string* s, uint32 value); +}; + +inline const char* Varint::Parse32WithLimit(const char* p, + const char* l, + uint32* OUTPUT) { + const unsigned char* ptr = reinterpret_cast(p); + const unsigned char* limit = reinterpret_cast(l); + uint32 b, result; + if (ptr >= limit) return NULL; + b = *(ptr++); result = b & 127; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 7; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 14; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 21; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 28; if (b < 16) goto done; + return NULL; // Value is too long to be a varint32 + done: + *OUTPUT = result; + return reinterpret_cast(ptr); +} + +inline char* Varint::Encode32(char* sptr, uint32 v) { + // Operate on characters as unsigneds + unsigned char* ptr = reinterpret_cast(sptr); + static const int B = 128; + if (v < (1<<7)) { + *(ptr++) = v; + } else if (v < (1<<14)) { + *(ptr++) = v | B; + *(ptr++) = v>>7; + } else if (v < (1<<21)) { + *(ptr++) = v | B; + *(ptr++) = (v>>7) | B; + *(ptr++) = v>>14; + } else if (v < (1<<28)) { + *(ptr++) = v | B; + *(ptr++) = (v>>7) | B; + *(ptr++) = (v>>14) | B; + *(ptr++) = v>>21; + } else { + *(ptr++) = v | B; + *(ptr++) = (v>>7) | B; + *(ptr++) = (v>>14) | B; + *(ptr++) = (v>>21) | B; + *(ptr++) = v>>28; + } + return reinterpret_cast(ptr); +} + +// If you know the internal layout of the std::string in use, you can +// replace this function with one that resizes the string without +// filling the new space with zeros (if applicable) -- +// it will be non-portable but faster. +inline void STLStringResizeUninitialized(string* s, size_t new_size) { + s->resize(new_size); +} + +// Return a mutable char* pointing to a string's internal buffer, +// which may not be null-terminated. Writing through this pointer will +// modify the string. +// +// string_as_array(&str)[i] is valid for 0 <= i < str.size() until the +// next call to a string method that invalidates iterators. +// +// As of 2006-04, there is no standard-blessed way of getting a +// mutable reference to a string's internal buffer. However, issue 530 +// (http://www.open-std.org/JTC1/SC22/WG21/docs/lwg-defects.html#530) +// proposes this as the method. It will officially be part of the standard +// for C++0x. This should already work on all current implementations. +inline char* string_as_array(string* str) { + return str->empty() ? NULL : &*str->begin(); +} + +} // namespace snappy + +#endif // THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_ diff --git a/external/snappy-1.1.3/snappy.cc b/external/snappy-1.1.3/snappy.cc new file mode 100644 index 0000000..a04c1ac --- /dev/null +++ b/external/snappy-1.1.3/snappy.cc @@ -0,0 +1,1553 @@ +// Copyright 2005 Google Inc. All Rights Reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#include "snappy/snappy.h" +#include "snappy-internal.h" +#include "snappy-sinksource.h" + +#include + +#include +#include +#include + + +namespace snappy { + +// Any hash function will produce a valid compressed bitstream, but a good +// hash function reduces the number of collisions and thus yields better +// compression for compressible input, and more speed for incompressible +// input. Of course, it doesn't hurt if the hash function is reasonably fast +// either, as it gets called a lot. +static inline uint32 HashBytes(uint32 bytes, int shift) { + uint32 kMul = 0x1e35a7bd; + return (bytes * kMul) >> shift; +} +static inline uint32 Hash(const char* p, int shift) { + return HashBytes(UNALIGNED_LOAD32(p), shift); +} + +size_t MaxCompressedLength(size_t source_len) { + // Compressed data can be defined as: + // compressed := item* literal* + // item := literal* copy + // + // The trailing literal sequence has a space blowup of at most 62/60 + // since a literal of length 60 needs one tag byte + one extra byte + // for length information. + // + // Item blowup is trickier to measure. Suppose the "copy" op copies + // 4 bytes of data. Because of a special check in the encoding code, + // we produce a 4-byte copy only if the offset is < 65536. Therefore + // the copy op takes 3 bytes to encode, and this type of item leads + // to at most the 62/60 blowup for representing literals. + // + // Suppose the "copy" op copies 5 bytes of data. If the offset is big + // enough, it will take 5 bytes to encode the copy op. Therefore the + // worst case here is a one-byte literal followed by a five-byte copy. + // I.e., 6 bytes of input turn into 7 bytes of "compressed" data. + // + // This last factor dominates the blowup, so the final estimate is: + return 32 + source_len + source_len/6; +} + +enum { + LITERAL = 0, + COPY_1_BYTE_OFFSET = 1, // 3 bit length + 3 bits of offset in opcode + COPY_2_BYTE_OFFSET = 2, + COPY_4_BYTE_OFFSET = 3 +}; +static const int kMaximumTagLength = 5; // COPY_4_BYTE_OFFSET plus the actual offset. + +// Copy "len" bytes from "src" to "op", one byte at a time. Used for +// handling COPY operations where the input and output regions may +// overlap. For example, suppose: +// src == "ab" +// op == src + 2 +// len == 20 +// After IncrementalCopy(src, op, len), the result will have +// eleven copies of "ab" +// ababababababababababab +// Note that this does not match the semantics of either memcpy() +// or memmove(). +static inline void IncrementalCopy(const char* src, char* op, ssize_t len) { + assert(len > 0); + do { + *op++ = *src++; + } while (--len > 0); +} + +// Equivalent to IncrementalCopy except that it can write up to ten extra +// bytes after the end of the copy, and that it is faster. +// +// The main part of this loop is a simple copy of eight bytes at a time until +// we've copied (at least) the requested amount of bytes. However, if op and +// src are less than eight bytes apart (indicating a repeating pattern of +// length < 8), we first need to expand the pattern in order to get the correct +// results. For instance, if the buffer looks like this, with the eight-byte +// and patterns marked as intervals: +// +// abxxxxxxxxxxxx +// [------] src +// [------] op +// +// a single eight-byte copy from to will repeat the pattern once, +// after which we can move two bytes without moving : +// +// ababxxxxxxxxxx +// [------] src +// [------] op +// +// and repeat the exercise until the two no longer overlap. +// +// This allows us to do very well in the special case of one single byte +// repeated many times, without taking a big hit for more general cases. +// +// The worst case of extra writing past the end of the match occurs when +// op - src == 1 and len == 1; the last copy will read from byte positions +// [0..7] and write to [4..11], whereas it was only supposed to write to +// position 1. Thus, ten excess bytes. + +namespace { + +const int kMaxIncrementCopyOverflow = 10; + +inline void IncrementalCopyFastPath(const char* src, char* op, ssize_t len) { + while (PREDICT_FALSE(op - src < 8)) { + UnalignedCopy64(src, op); + len -= op - src; + op += op - src; + } + while (len > 0) { + UnalignedCopy64(src, op); + src += 8; + op += 8; + len -= 8; + } +} + +} // namespace + +static inline char* EmitLiteral(char* op, + const char* literal, + int len, + bool allow_fast_path) { + int n = len - 1; // Zero-length literals are disallowed + if (n < 60) { + // Fits in tag byte + *op++ = LITERAL | (n << 2); + + // The vast majority of copies are below 16 bytes, for which a + // call to memcpy is overkill. This fast path can sometimes + // copy up to 15 bytes too much, but that is okay in the + // main loop, since we have a bit to go on for both sides: + // + // - The input will always have kInputMarginBytes = 15 extra + // available bytes, as long as we're in the main loop, and + // if not, allow_fast_path = false. + // - The output will always have 32 spare bytes (see + // MaxCompressedLength). + if (allow_fast_path && len <= 16) { + UnalignedCopy64(literal, op); + UnalignedCopy64(literal + 8, op + 8); + return op + len; + } + } else { + // Encode in upcoming bytes + char* base = op; + int count = 0; + op++; + while (n > 0) { + *op++ = n & 0xff; + n >>= 8; + count++; + } + assert(count >= 1); + assert(count <= 4); + *base = LITERAL | ((59+count) << 2); + } + memcpy(op, literal, len); + return op + len; +} + +static inline char* EmitCopyLessThan64(char* op, size_t offset, int len) { + assert(len <= 64); + assert(len >= 4); + assert(offset < 65536); + + if ((len < 12) && (offset < 2048)) { + size_t len_minus_4 = len - 4; + assert(len_minus_4 < 8); // Must fit in 3 bits + *op++ = COPY_1_BYTE_OFFSET + ((len_minus_4) << 2) + ((offset >> 8) << 5); + *op++ = offset & 0xff; + } else { + *op++ = COPY_2_BYTE_OFFSET + ((len-1) << 2); + LittleEndian::Store16(op, offset); + op += 2; + } + return op; +} + +static inline char* EmitCopy(char* op, size_t offset, int len) { + // Emit 64 byte copies but make sure to keep at least four bytes reserved + while (PREDICT_FALSE(len >= 68)) { + op = EmitCopyLessThan64(op, offset, 64); + len -= 64; + } + + // Emit an extra 60 byte copy if have too much data to fit in one copy + if (len > 64) { + op = EmitCopyLessThan64(op, offset, 60); + len -= 60; + } + + // Emit remainder + op = EmitCopyLessThan64(op, offset, len); + return op; +} + + +bool GetUncompressedLength(const char* start, size_t n, size_t* result) { + uint32 v = 0; + const char* limit = start + n; + if (Varint::Parse32WithLimit(start, limit, &v) != NULL) { + *result = v; + return true; + } else { + return false; + } +} + +namespace internal { +uint16* WorkingMemory::GetHashTable(size_t input_size, int* table_size) { + // Use smaller hash table when input.size() is smaller, since we + // fill the table, incurring O(hash table size) overhead for + // compression, and if the input is short, we won't need that + // many hash table entries anyway. + assert(kMaxHashTableSize >= 256); + size_t htsize = 256; + while (htsize < kMaxHashTableSize && htsize < input_size) { + htsize <<= 1; + } + + uint16* table; + if (htsize <= ARRAYSIZE(small_table_)) { + table = small_table_; + } else { + if (large_table_ == NULL) { + large_table_ = new uint16[kMaxHashTableSize]; + } + table = large_table_; + } + + *table_size = htsize; + memset(table, 0, htsize * sizeof(*table)); + return table; +} +} // end namespace internal + +// For 0 <= offset <= 4, GetUint32AtOffset(GetEightBytesAt(p), offset) will +// equal UNALIGNED_LOAD32(p + offset). Motivation: On x86-64 hardware we have +// empirically found that overlapping loads such as +// UNALIGNED_LOAD32(p) ... UNALIGNED_LOAD32(p+1) ... UNALIGNED_LOAD32(p+2) +// are slower than UNALIGNED_LOAD64(p) followed by shifts and casts to uint32. +// +// We have different versions for 64- and 32-bit; ideally we would avoid the +// two functions and just inline the UNALIGNED_LOAD64 call into +// GetUint32AtOffset, but GCC (at least not as of 4.6) is seemingly not clever +// enough to avoid loading the value multiple times then. For 64-bit, the load +// is done when GetEightBytesAt() is called, whereas for 32-bit, the load is +// done at GetUint32AtOffset() time. + +#ifdef ARCH_K8 + +typedef uint64 EightBytesReference; + +static inline EightBytesReference GetEightBytesAt(const char* ptr) { + return UNALIGNED_LOAD64(ptr); +} + +static inline uint32 GetUint32AtOffset(uint64 v, int offset) { + assert(offset >= 0); + assert(offset <= 4); + return v >> (LittleEndian::IsLittleEndian() ? 8 * offset : 32 - 8 * offset); +} + +#else + +typedef const char* EightBytesReference; + +static inline EightBytesReference GetEightBytesAt(const char* ptr) { + return ptr; +} + +static inline uint32 GetUint32AtOffset(const char* v, int offset) { + assert(offset >= 0); + assert(offset <= 4); + return UNALIGNED_LOAD32(v + offset); +} + +#endif + +// Flat array compression that does not emit the "uncompressed length" +// prefix. Compresses "input" string to the "*op" buffer. +// +// REQUIRES: "input" is at most "kBlockSize" bytes long. +// REQUIRES: "op" points to an array of memory that is at least +// "MaxCompressedLength(input.size())" in size. +// REQUIRES: All elements in "table[0..table_size-1]" are initialized to zero. +// REQUIRES: "table_size" is a power of two +// +// Returns an "end" pointer into "op" buffer. +// "end - op" is the compressed size of "input". +namespace internal { +char* CompressFragment(const char* input, + size_t input_size, + char* op, + uint16* table, + const int table_size) { + // "ip" is the input pointer, and "op" is the output pointer. + const char* ip = input; + assert(input_size <= kBlockSize); + assert((table_size & (table_size - 1)) == 0); // table must be power of two + const int shift = 32 - Bits::Log2Floor(table_size); + assert(static_cast(kuint32max >> shift) == table_size - 1); + const char* ip_end = input + input_size; + const char* base_ip = ip; + // Bytes in [next_emit, ip) will be emitted as literal bytes. Or + // [next_emit, ip_end) after the main loop. + const char* next_emit = ip; + + const size_t kInputMarginBytes = 15; + if (PREDICT_TRUE(input_size >= kInputMarginBytes)) { + const char* ip_limit = input + input_size - kInputMarginBytes; + + for (uint32 next_hash = Hash(++ip, shift); ; ) { + assert(next_emit < ip); + // The body of this loop calls EmitLiteral once and then EmitCopy one or + // more times. (The exception is that when we're close to exhausting + // the input we goto emit_remainder.) + // + // In the first iteration of this loop we're just starting, so + // there's nothing to copy, so calling EmitLiteral once is + // necessary. And we only start a new iteration when the + // current iteration has determined that a call to EmitLiteral will + // precede the next call to EmitCopy (if any). + // + // Step 1: Scan forward in the input looking for a 4-byte-long match. + // If we get close to exhausting the input then goto emit_remainder. + // + // Heuristic match skipping: If 32 bytes are scanned with no matches + // found, start looking only at every other byte. If 32 more bytes are + // scanned, look at every third byte, etc.. When a match is found, + // immediately go back to looking at every byte. This is a small loss + // (~5% performance, ~0.1% density) for compressible data due to more + // bookkeeping, but for non-compressible data (such as JPEG) it's a huge + // win since the compressor quickly "realizes" the data is incompressible + // and doesn't bother looking for matches everywhere. + // + // The "skip" variable keeps track of how many bytes there are since the + // last match; dividing it by 32 (ie. right-shifting by five) gives the + // number of bytes to move ahead for each iteration. + uint32 skip = 32; + + const char* next_ip = ip; + const char* candidate; + do { + ip = next_ip; + uint32 hash = next_hash; + assert(hash == Hash(ip, shift)); + uint32 bytes_between_hash_lookups = skip++ >> 5; + next_ip = ip + bytes_between_hash_lookups; + if (PREDICT_FALSE(next_ip > ip_limit)) { + goto emit_remainder; + } + next_hash = Hash(next_ip, shift); + candidate = base_ip + table[hash]; + assert(candidate >= base_ip); + assert(candidate < ip); + + table[hash] = ip - base_ip; + } while (PREDICT_TRUE(UNALIGNED_LOAD32(ip) != + UNALIGNED_LOAD32(candidate))); + + // Step 2: A 4-byte match has been found. We'll later see if more + // than 4 bytes match. But, prior to the match, input + // bytes [next_emit, ip) are unmatched. Emit them as "literal bytes." + assert(next_emit + 16 <= ip_end); + op = EmitLiteral(op, next_emit, ip - next_emit, true); + + // Step 3: Call EmitCopy, and then see if another EmitCopy could + // be our next move. Repeat until we find no match for the + // input immediately after what was consumed by the last EmitCopy call. + // + // If we exit this loop normally then we need to call EmitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can exit + // this loop via goto if we get close to exhausting the input. + EightBytesReference input_bytes; + uint32 candidate_bytes = 0; + + do { + // We have a 4-byte match at ip, and no need to emit any + // "literal bytes" prior to ip. + const char* base = ip; + int matched = 4 + FindMatchLength(candidate + 4, ip + 4, ip_end); + ip += matched; + size_t offset = base - candidate; + assert(0 == memcmp(base, candidate, matched)); + op = EmitCopy(op, offset, matched); + // We could immediately start working at ip now, but to improve + // compression we first update table[Hash(ip - 1, ...)]. + const char* insert_tail = ip - 1; + next_emit = ip; + if (PREDICT_FALSE(ip >= ip_limit)) { + goto emit_remainder; + } + input_bytes = GetEightBytesAt(insert_tail); + uint32 prev_hash = HashBytes(GetUint32AtOffset(input_bytes, 0), shift); + table[prev_hash] = ip - base_ip - 1; + uint32 cur_hash = HashBytes(GetUint32AtOffset(input_bytes, 1), shift); + candidate = base_ip + table[cur_hash]; + candidate_bytes = UNALIGNED_LOAD32(candidate); + table[cur_hash] = ip - base_ip; + } while (GetUint32AtOffset(input_bytes, 1) == candidate_bytes); + + next_hash = HashBytes(GetUint32AtOffset(input_bytes, 2), shift); + ++ip; + } + } + + emit_remainder: + // Emit the remaining bytes as a literal + if (next_emit < ip_end) { + op = EmitLiteral(op, next_emit, ip_end - next_emit, false); + } + + return op; +} +} // end namespace internal + +// Signature of output types needed by decompression code. +// The decompression code is templatized on a type that obeys this +// signature so that we do not pay virtual function call overhead in +// the middle of a tight decompression loop. +// +// class DecompressionWriter { +// public: +// // Called before decompression +// void SetExpectedLength(size_t length); +// +// // Called after decompression +// bool CheckLength() const; +// +// // Called repeatedly during decompression +// bool Append(const char* ip, size_t length); +// bool AppendFromSelf(uint32 offset, size_t length); +// +// // The rules for how TryFastAppend differs from Append are somewhat +// // convoluted: +// // +// // - TryFastAppend is allowed to decline (return false) at any +// // time, for any reason -- just "return false" would be +// // a perfectly legal implementation of TryFastAppend. +// // The intention is for TryFastAppend to allow a fast path +// // in the common case of a small append. +// // - TryFastAppend is allowed to read up to bytes +// // from the input buffer, whereas Append is allowed to read +// // . However, if it returns true, it must leave +// // at least five (kMaximumTagLength) bytes in the input buffer +// // afterwards, so that there is always enough space to read the +// // next tag without checking for a refill. +// // - TryFastAppend must always return decline (return false) +// // if is 61 or more, as in this case the literal length is not +// // decoded fully. In practice, this should not be a big problem, +// // as it is unlikely that one would implement a fast path accepting +// // this much data. +// // +// bool TryFastAppend(const char* ip, size_t available, size_t length); +// }; + +// ----------------------------------------------------------------------- +// Lookup table for decompression code. Generated by ComputeTable() below. +// ----------------------------------------------------------------------- + +// Mapping from i in range [0,4] to a mask to extract the bottom 8*i bits +static const uint32 wordmask[] = { + 0u, 0xffu, 0xffffu, 0xffffffu, 0xffffffffu +}; + +// Data stored per entry in lookup table: +// Range Bits-used Description +// ------------------------------------ +// 1..64 0..7 Literal/copy length encoded in opcode byte +// 0..7 8..10 Copy offset encoded in opcode byte / 256 +// 0..4 11..13 Extra bytes after opcode +// +// We use eight bits for the length even though 7 would have sufficed +// because of efficiency reasons: +// (1) Extracting a byte is faster than a bit-field +// (2) It properly aligns copy offset so we do not need a <<8 +static const uint16 char_table[256] = { + 0x0001, 0x0804, 0x1001, 0x2001, 0x0002, 0x0805, 0x1002, 0x2002, + 0x0003, 0x0806, 0x1003, 0x2003, 0x0004, 0x0807, 0x1004, 0x2004, + 0x0005, 0x0808, 0x1005, 0x2005, 0x0006, 0x0809, 0x1006, 0x2006, + 0x0007, 0x080a, 0x1007, 0x2007, 0x0008, 0x080b, 0x1008, 0x2008, + 0x0009, 0x0904, 0x1009, 0x2009, 0x000a, 0x0905, 0x100a, 0x200a, + 0x000b, 0x0906, 0x100b, 0x200b, 0x000c, 0x0907, 0x100c, 0x200c, + 0x000d, 0x0908, 0x100d, 0x200d, 0x000e, 0x0909, 0x100e, 0x200e, + 0x000f, 0x090a, 0x100f, 0x200f, 0x0010, 0x090b, 0x1010, 0x2010, + 0x0011, 0x0a04, 0x1011, 0x2011, 0x0012, 0x0a05, 0x1012, 0x2012, + 0x0013, 0x0a06, 0x1013, 0x2013, 0x0014, 0x0a07, 0x1014, 0x2014, + 0x0015, 0x0a08, 0x1015, 0x2015, 0x0016, 0x0a09, 0x1016, 0x2016, + 0x0017, 0x0a0a, 0x1017, 0x2017, 0x0018, 0x0a0b, 0x1018, 0x2018, + 0x0019, 0x0b04, 0x1019, 0x2019, 0x001a, 0x0b05, 0x101a, 0x201a, + 0x001b, 0x0b06, 0x101b, 0x201b, 0x001c, 0x0b07, 0x101c, 0x201c, + 0x001d, 0x0b08, 0x101d, 0x201d, 0x001e, 0x0b09, 0x101e, 0x201e, + 0x001f, 0x0b0a, 0x101f, 0x201f, 0x0020, 0x0b0b, 0x1020, 0x2020, + 0x0021, 0x0c04, 0x1021, 0x2021, 0x0022, 0x0c05, 0x1022, 0x2022, + 0x0023, 0x0c06, 0x1023, 0x2023, 0x0024, 0x0c07, 0x1024, 0x2024, + 0x0025, 0x0c08, 0x1025, 0x2025, 0x0026, 0x0c09, 0x1026, 0x2026, + 0x0027, 0x0c0a, 0x1027, 0x2027, 0x0028, 0x0c0b, 0x1028, 0x2028, + 0x0029, 0x0d04, 0x1029, 0x2029, 0x002a, 0x0d05, 0x102a, 0x202a, + 0x002b, 0x0d06, 0x102b, 0x202b, 0x002c, 0x0d07, 0x102c, 0x202c, + 0x002d, 0x0d08, 0x102d, 0x202d, 0x002e, 0x0d09, 0x102e, 0x202e, + 0x002f, 0x0d0a, 0x102f, 0x202f, 0x0030, 0x0d0b, 0x1030, 0x2030, + 0x0031, 0x0e04, 0x1031, 0x2031, 0x0032, 0x0e05, 0x1032, 0x2032, + 0x0033, 0x0e06, 0x1033, 0x2033, 0x0034, 0x0e07, 0x1034, 0x2034, + 0x0035, 0x0e08, 0x1035, 0x2035, 0x0036, 0x0e09, 0x1036, 0x2036, + 0x0037, 0x0e0a, 0x1037, 0x2037, 0x0038, 0x0e0b, 0x1038, 0x2038, + 0x0039, 0x0f04, 0x1039, 0x2039, 0x003a, 0x0f05, 0x103a, 0x203a, + 0x003b, 0x0f06, 0x103b, 0x203b, 0x003c, 0x0f07, 0x103c, 0x203c, + 0x0801, 0x0f08, 0x103d, 0x203d, 0x1001, 0x0f09, 0x103e, 0x203e, + 0x1801, 0x0f0a, 0x103f, 0x203f, 0x2001, 0x0f0b, 0x1040, 0x2040 +}; + +// In debug mode, allow optional computation of the table at startup. +// Also, check that the decompression table is correct. +#ifndef NDEBUG +DEFINE_bool(snappy_dump_decompression_table, false, + "If true, we print the decompression table at startup."); + +static uint16 MakeEntry(unsigned int extra, + unsigned int len, + unsigned int copy_offset) { + // Check that all of the fields fit within the allocated space + assert(extra == (extra & 0x7)); // At most 3 bits + assert(copy_offset == (copy_offset & 0x7)); // At most 3 bits + assert(len == (len & 0x7f)); // At most 7 bits + return len | (copy_offset << 8) | (extra << 11); +} + +static void ComputeTable() { + uint16 dst[256]; + + // Place invalid entries in all places to detect missing initialization + int assigned = 0; + for (int i = 0; i < 256; i++) { + dst[i] = 0xffff; + } + + // Small LITERAL entries. We store (len-1) in the top 6 bits. + for (unsigned int len = 1; len <= 60; len++) { + dst[LITERAL | ((len-1) << 2)] = MakeEntry(0, len, 0); + assigned++; + } + + // Large LITERAL entries. We use 60..63 in the high 6 bits to + // encode the number of bytes of length info that follow the opcode. + for (unsigned int extra_bytes = 1; extra_bytes <= 4; extra_bytes++) { + // We set the length field in the lookup table to 1 because extra + // bytes encode len-1. + dst[LITERAL | ((extra_bytes+59) << 2)] = MakeEntry(extra_bytes, 1, 0); + assigned++; + } + + // COPY_1_BYTE_OFFSET. + // + // The tag byte in the compressed data stores len-4 in 3 bits, and + // offset/256 in 5 bits. offset%256 is stored in the next byte. + // + // This format is used for length in range [4..11] and offset in + // range [0..2047] + for (unsigned int len = 4; len < 12; len++) { + for (unsigned int offset = 0; offset < 2048; offset += 256) { + dst[COPY_1_BYTE_OFFSET | ((len-4)<<2) | ((offset>>8)<<5)] = + MakeEntry(1, len, offset>>8); + assigned++; + } + } + + // COPY_2_BYTE_OFFSET. + // Tag contains len-1 in top 6 bits, and offset in next two bytes. + for (unsigned int len = 1; len <= 64; len++) { + dst[COPY_2_BYTE_OFFSET | ((len-1)<<2)] = MakeEntry(2, len, 0); + assigned++; + } + + // COPY_4_BYTE_OFFSET. + // Tag contents len-1 in top 6 bits, and offset in next four bytes. + for (unsigned int len = 1; len <= 64; len++) { + dst[COPY_4_BYTE_OFFSET | ((len-1)<<2)] = MakeEntry(4, len, 0); + assigned++; + } + + // Check that each entry was initialized exactly once. + if (assigned != 256) { + fprintf(stderr, "ComputeTable: assigned only %d of 256\n", assigned); + abort(); + } + for (int i = 0; i < 256; i++) { + if (dst[i] == 0xffff) { + fprintf(stderr, "ComputeTable: did not assign byte %d\n", i); + abort(); + } + } + + if (FLAGS_snappy_dump_decompression_table) { + printf("static const uint16 char_table[256] = {\n "); + for (int i = 0; i < 256; i++) { + printf("0x%04x%s", + dst[i], + ((i == 255) ? "\n" : (((i%8) == 7) ? ",\n " : ", "))); + } + printf("};\n"); + } + + // Check that computed table matched recorded table + for (int i = 0; i < 256; i++) { + if (dst[i] != char_table[i]) { + fprintf(stderr, "ComputeTable: byte %d: computed (%x), expect (%x)\n", + i, static_cast(dst[i]), static_cast(char_table[i])); + abort(); + } + } +} +#endif /* !NDEBUG */ + +// Helper class for decompression +class SnappyDecompressor { + private: + Source* reader_; // Underlying source of bytes to decompress + const char* ip_; // Points to next buffered byte + const char* ip_limit_; // Points just past buffered bytes + uint32 peeked_; // Bytes peeked from reader (need to skip) + bool eof_; // Hit end of input without an error? + char scratch_[kMaximumTagLength]; // See RefillTag(). + + // Ensure that all of the tag metadata for the next tag is available + // in [ip_..ip_limit_-1]. Also ensures that [ip,ip+4] is readable even + // if (ip_limit_ - ip_ < 5). + // + // Returns true on success, false on error or end of input. + bool RefillTag(); + + public: + explicit SnappyDecompressor(Source* reader) + : reader_(reader), + ip_(NULL), + ip_limit_(NULL), + peeked_(0), + eof_(false) { + } + + ~SnappyDecompressor() { + // Advance past any bytes we peeked at from the reader + reader_->Skip(peeked_); + } + + // Returns true iff we have hit the end of the input without an error. + bool eof() const { + return eof_; + } + + // Read the uncompressed length stored at the start of the compressed data. + // On succcess, stores the length in *result and returns true. + // On failure, returns false. + bool ReadUncompressedLength(uint32* result) { + assert(ip_ == NULL); // Must not have read anything yet + // Length is encoded in 1..5 bytes + *result = 0; + uint32 shift = 0; + while (true) { + if (shift >= 32) return false; + size_t n; + const char* ip = reader_->Peek(&n); + if (n == 0) return false; + const unsigned char c = *(reinterpret_cast(ip)); + reader_->Skip(1); + *result |= static_cast(c & 0x7f) << shift; + if (c < 128) { + break; + } + shift += 7; + } + return true; + } + + // Process the next item found in the input. + // Returns true if successful, false on error or end of input. + template + void DecompressAllTags(Writer* writer) { + const char* ip = ip_; + + // We could have put this refill fragment only at the beginning of the loop. + // However, duplicating it at the end of each branch gives the compiler more + // scope to optimize the expression based on the local + // context, which overall increases speed. + #define MAYBE_REFILL() \ + if (ip_limit_ - ip < kMaximumTagLength) { \ + ip_ = ip; \ + if (!RefillTag()) return; \ + ip = ip_; \ + } + + MAYBE_REFILL(); + for ( ;; ) { + const unsigned char c = *(reinterpret_cast(ip++)); + + if ((c & 0x3) == LITERAL) { + size_t literal_length = (c >> 2) + 1u; + if (writer->TryFastAppend(ip, ip_limit_ - ip, literal_length)) { + assert(literal_length < 61); + ip += literal_length; + // NOTE(user): There is no MAYBE_REFILL() here, as TryFastAppend() + // will not return true unless there's already at least five spare + // bytes in addition to the literal. + continue; + } + if (PREDICT_FALSE(literal_length >= 61)) { + // Long literal. + const size_t literal_length_length = literal_length - 60; + literal_length = + (LittleEndian::Load32(ip) & wordmask[literal_length_length]) + 1; + ip += literal_length_length; + } + + size_t avail = ip_limit_ - ip; + while (avail < literal_length) { + if (!writer->Append(ip, avail)) return; + literal_length -= avail; + reader_->Skip(peeked_); + size_t n; + ip = reader_->Peek(&n); + avail = n; + peeked_ = avail; + if (avail == 0) return; // Premature end of input + ip_limit_ = ip + avail; + } + if (!writer->Append(ip, literal_length)) { + return; + } + ip += literal_length; + MAYBE_REFILL(); + } else { + const uint32 entry = char_table[c]; + const uint32 trailer = LittleEndian::Load32(ip) & wordmask[entry >> 11]; + const uint32 length = entry & 0xff; + ip += entry >> 11; + + // copy_offset/256 is encoded in bits 8..10. By just fetching + // those bits, we get copy_offset (since the bit-field starts at + // bit 8). + const uint32 copy_offset = entry & 0x700; + if (!writer->AppendFromSelf(copy_offset + trailer, length)) { + return; + } + MAYBE_REFILL(); + } + } + +#undef MAYBE_REFILL + } +}; + +bool SnappyDecompressor::RefillTag() { + const char* ip = ip_; + if (ip == ip_limit_) { + // Fetch a new fragment from the reader + reader_->Skip(peeked_); // All peeked bytes are used up + size_t n; + ip = reader_->Peek(&n); + peeked_ = n; + if (n == 0) { + eof_ = true; + return false; + } + ip_limit_ = ip + n; + } + + // Read the tag character + assert(ip < ip_limit_); + const unsigned char c = *(reinterpret_cast(ip)); + const uint32 entry = char_table[c]; + const uint32 needed = (entry >> 11) + 1; // +1 byte for 'c' + assert(needed <= sizeof(scratch_)); + + // Read more bytes from reader if needed + uint32 nbuf = ip_limit_ - ip; + if (nbuf < needed) { + // Stitch together bytes from ip and reader to form the word + // contents. We store the needed bytes in "scratch_". They + // will be consumed immediately by the caller since we do not + // read more than we need. + memmove(scratch_, ip, nbuf); + reader_->Skip(peeked_); // All peeked bytes are used up + peeked_ = 0; + while (nbuf < needed) { + size_t length; + const char* src = reader_->Peek(&length); + if (length == 0) return false; + uint32 to_add = min(needed - nbuf, length); + memcpy(scratch_ + nbuf, src, to_add); + nbuf += to_add; + reader_->Skip(to_add); + } + assert(nbuf == needed); + ip_ = scratch_; + ip_limit_ = scratch_ + needed; + } else if (nbuf < kMaximumTagLength) { + // Have enough bytes, but move into scratch_ so that we do not + // read past end of input + memmove(scratch_, ip, nbuf); + reader_->Skip(peeked_); // All peeked bytes are used up + peeked_ = 0; + ip_ = scratch_; + ip_limit_ = scratch_ + nbuf; + } else { + // Pass pointer to buffer returned by reader_. + ip_ = ip; + } + return true; +} + +template +static bool InternalUncompress(Source* r, Writer* writer) { + // Read the uncompressed length from the front of the compressed input + SnappyDecompressor decompressor(r); + uint32 uncompressed_len = 0; + if (!decompressor.ReadUncompressedLength(&uncompressed_len)) return false; + return InternalUncompressAllTags(&decompressor, writer, uncompressed_len); +} + +template +static bool InternalUncompressAllTags(SnappyDecompressor* decompressor, + Writer* writer, + uint32 uncompressed_len) { + writer->SetExpectedLength(uncompressed_len); + + // Process the entire input + decompressor->DecompressAllTags(writer); + writer->Flush(); + return (decompressor->eof() && writer->CheckLength()); +} + +bool GetUncompressedLength(Source* source, uint32* result) { + SnappyDecompressor decompressor(source); + return decompressor.ReadUncompressedLength(result); +} + +size_t Compress(Source* reader, Sink* writer) { + size_t written = 0; + size_t N = reader->Available(); + char ulength[Varint::kMax32]; + char* p = Varint::Encode32(ulength, N); + writer->Append(ulength, p-ulength); + written += (p - ulength); + + internal::WorkingMemory wmem; + char* scratch = NULL; + char* scratch_output = NULL; + + while (N > 0) { + // Get next block to compress (without copying if possible) + size_t fragment_size; + const char* fragment = reader->Peek(&fragment_size); + assert(fragment_size != 0); // premature end of input + const size_t num_to_read = min(N, kBlockSize); + size_t bytes_read = fragment_size; + + size_t pending_advance = 0; + if (bytes_read >= num_to_read) { + // Buffer returned by reader is large enough + pending_advance = num_to_read; + fragment_size = num_to_read; + } else { + // Read into scratch buffer + if (scratch == NULL) { + // If this is the last iteration, we want to allocate N bytes + // of space, otherwise the max possible kBlockSize space. + // num_to_read contains exactly the correct value + scratch = new char[num_to_read]; + } + memcpy(scratch, fragment, bytes_read); + reader->Skip(bytes_read); + + while (bytes_read < num_to_read) { + fragment = reader->Peek(&fragment_size); + size_t n = min(fragment_size, num_to_read - bytes_read); + memcpy(scratch + bytes_read, fragment, n); + bytes_read += n; + reader->Skip(n); + } + assert(bytes_read == num_to_read); + fragment = scratch; + fragment_size = num_to_read; + } + assert(fragment_size == num_to_read); + + // Get encoding table for compression + int table_size; + uint16* table = wmem.GetHashTable(num_to_read, &table_size); + + // Compress input_fragment and append to dest + const int max_output = MaxCompressedLength(num_to_read); + + // Need a scratch buffer for the output, in case the byte sink doesn't + // have room for us directly. + if (scratch_output == NULL) { + scratch_output = new char[max_output]; + } else { + // Since we encode kBlockSize regions followed by a region + // which is <= kBlockSize in length, a previously allocated + // scratch_output[] region is big enough for this iteration. + } + char* dest = writer->GetAppendBuffer(max_output, scratch_output); + char* end = internal::CompressFragment(fragment, fragment_size, + dest, table, table_size); + writer->Append(dest, end - dest); + written += (end - dest); + + N -= num_to_read; + reader->Skip(pending_advance); + } + + delete[] scratch; + delete[] scratch_output; + + return written; +} + +// ----------------------------------------------------------------------- +// IOVec interfaces +// ----------------------------------------------------------------------- + +// A type that writes to an iovec. +// Note that this is not a "ByteSink", but a type that matches the +// Writer template argument to SnappyDecompressor::DecompressAllTags(). +class SnappyIOVecWriter { + private: + const struct iovec* output_iov_; + const size_t output_iov_count_; + + // We are currently writing into output_iov_[curr_iov_index_]. + int curr_iov_index_; + + // Bytes written to output_iov_[curr_iov_index_] so far. + size_t curr_iov_written_; + + // Total bytes decompressed into output_iov_ so far. + size_t total_written_; + + // Maximum number of bytes that will be decompressed into output_iov_. + size_t output_limit_; + + inline char* GetIOVecPointer(int index, size_t offset) { + return reinterpret_cast(output_iov_[index].iov_base) + + offset; + } + + public: + // Does not take ownership of iov. iov must be valid during the + // entire lifetime of the SnappyIOVecWriter. + inline SnappyIOVecWriter(const struct iovec* iov, size_t iov_count) + : output_iov_(iov), + output_iov_count_(iov_count), + curr_iov_index_(0), + curr_iov_written_(0), + total_written_(0), + output_limit_(-1) { + } + + inline void SetExpectedLength(size_t len) { + output_limit_ = len; + } + + inline bool CheckLength() const { + return total_written_ == output_limit_; + } + + inline bool Append(const char* ip, size_t len) { + if (total_written_ + len > output_limit_) { + return false; + } + + while (len > 0) { + assert(curr_iov_written_ <= output_iov_[curr_iov_index_].iov_len); + if (curr_iov_written_ >= output_iov_[curr_iov_index_].iov_len) { + // This iovec is full. Go to the next one. + if (curr_iov_index_ + 1 >= output_iov_count_) { + return false; + } + curr_iov_written_ = 0; + ++curr_iov_index_; + } + + const size_t to_write = std::min( + len, output_iov_[curr_iov_index_].iov_len - curr_iov_written_); + memcpy(GetIOVecPointer(curr_iov_index_, curr_iov_written_), + ip, + to_write); + curr_iov_written_ += to_write; + total_written_ += to_write; + ip += to_write; + len -= to_write; + } + + return true; + } + + inline bool TryFastAppend(const char* ip, size_t available, size_t len) { + const size_t space_left = output_limit_ - total_written_; + if (len <= 16 && available >= 16 + kMaximumTagLength && space_left >= 16 && + output_iov_[curr_iov_index_].iov_len - curr_iov_written_ >= 16) { + // Fast path, used for the majority (about 95%) of invocations. + char* ptr = GetIOVecPointer(curr_iov_index_, curr_iov_written_); + UnalignedCopy64(ip, ptr); + UnalignedCopy64(ip + 8, ptr + 8); + curr_iov_written_ += len; + total_written_ += len; + return true; + } + + return false; + } + + inline bool AppendFromSelf(size_t offset, size_t len) { + if (offset > total_written_ || offset == 0) { + return false; + } + const size_t space_left = output_limit_ - total_written_; + if (len > space_left) { + return false; + } + + // Locate the iovec from which we need to start the copy. + int from_iov_index = curr_iov_index_; + size_t from_iov_offset = curr_iov_written_; + while (offset > 0) { + if (from_iov_offset >= offset) { + from_iov_offset -= offset; + break; + } + + offset -= from_iov_offset; + --from_iov_index; + assert(from_iov_index >= 0); + from_iov_offset = output_iov_[from_iov_index].iov_len; + } + + // Copy bytes starting from the iovec pointed to by from_iov_index to + // the current iovec. + while (len > 0) { + assert(from_iov_index <= curr_iov_index_); + if (from_iov_index != curr_iov_index_) { + const size_t to_copy = std::min( + output_iov_[from_iov_index].iov_len - from_iov_offset, + len); + Append(GetIOVecPointer(from_iov_index, from_iov_offset), to_copy); + len -= to_copy; + if (len > 0) { + ++from_iov_index; + from_iov_offset = 0; + } + } else { + assert(curr_iov_written_ <= output_iov_[curr_iov_index_].iov_len); + size_t to_copy = std::min(output_iov_[curr_iov_index_].iov_len - + curr_iov_written_, + len); + if (to_copy == 0) { + // This iovec is full. Go to the next one. + if (curr_iov_index_ + 1 >= output_iov_count_) { + return false; + } + ++curr_iov_index_; + curr_iov_written_ = 0; + continue; + } + if (to_copy > len) { + to_copy = len; + } + IncrementalCopy(GetIOVecPointer(from_iov_index, from_iov_offset), + GetIOVecPointer(curr_iov_index_, curr_iov_written_), + to_copy); + curr_iov_written_ += to_copy; + from_iov_offset += to_copy; + total_written_ += to_copy; + len -= to_copy; + } + } + + return true; + } + + inline void Flush() {} +}; + +bool RawUncompressToIOVec(const char* compressed, size_t compressed_length, + const struct iovec* iov, size_t iov_cnt) { + ByteArraySource reader(compressed, compressed_length); + return RawUncompressToIOVec(&reader, iov, iov_cnt); +} + +bool RawUncompressToIOVec(Source* compressed, const struct iovec* iov, + size_t iov_cnt) { + SnappyIOVecWriter output(iov, iov_cnt); + return InternalUncompress(compressed, &output); +} + +// ----------------------------------------------------------------------- +// Flat array interfaces +// ----------------------------------------------------------------------- + +// A type that writes to a flat array. +// Note that this is not a "ByteSink", but a type that matches the +// Writer template argument to SnappyDecompressor::DecompressAllTags(). +class SnappyArrayWriter { + private: + char* base_; + char* op_; + char* op_limit_; + + public: + inline explicit SnappyArrayWriter(char* dst) + : base_(dst), + op_(dst), + op_limit_(dst) { + } + + inline void SetExpectedLength(size_t len) { + op_limit_ = op_ + len; + } + + inline bool CheckLength() const { + return op_ == op_limit_; + } + + inline bool Append(const char* ip, size_t len) { + char* op = op_; + const size_t space_left = op_limit_ - op; + if (space_left < len) { + return false; + } + memcpy(op, ip, len); + op_ = op + len; + return true; + } + + inline bool TryFastAppend(const char* ip, size_t available, size_t len) { + char* op = op_; + const size_t space_left = op_limit_ - op; + if (len <= 16 && available >= 16 + kMaximumTagLength && space_left >= 16) { + // Fast path, used for the majority (about 95%) of invocations. + UnalignedCopy64(ip, op); + UnalignedCopy64(ip + 8, op + 8); + op_ = op + len; + return true; + } else { + return false; + } + } + + inline bool AppendFromSelf(size_t offset, size_t len) { + char* op = op_; + const size_t space_left = op_limit_ - op; + + // Check if we try to append from before the start of the buffer. + // Normally this would just be a check for "produced < offset", + // but "produced <= offset - 1u" is equivalent for every case + // except the one where offset==0, where the right side will wrap around + // to a very big number. This is convenient, as offset==0 is another + // invalid case that we also want to catch, so that we do not go + // into an infinite loop. + assert(op >= base_); + size_t produced = op - base_; + if (produced <= offset - 1u) { + return false; + } + if (len <= 16 && offset >= 8 && space_left >= 16) { + // Fast path, used for the majority (70-80%) of dynamic invocations. + UnalignedCopy64(op - offset, op); + UnalignedCopy64(op - offset + 8, op + 8); + } else { + if (space_left >= len + kMaxIncrementCopyOverflow) { + IncrementalCopyFastPath(op - offset, op, len); + } else { + if (space_left < len) { + return false; + } + IncrementalCopy(op - offset, op, len); + } + } + + op_ = op + len; + return true; + } + inline size_t Produced() const { + return op_ - base_; + } + inline void Flush() {} +}; + +bool RawUncompress(const char* compressed, size_t n, char* uncompressed) { + ByteArraySource reader(compressed, n); + return RawUncompress(&reader, uncompressed); +} + +bool RawUncompress(Source* compressed, char* uncompressed) { + SnappyArrayWriter output(uncompressed); + return InternalUncompress(compressed, &output); +} + +bool Uncompress(const char* compressed, size_t n, string* uncompressed) { + size_t ulength; + if (!GetUncompressedLength(compressed, n, &ulength)) { + return false; + } + // On 32-bit builds: max_size() < kuint32max. Check for that instead + // of crashing (e.g., consider externally specified compressed data). + if (ulength > uncompressed->max_size()) { + return false; + } + STLStringResizeUninitialized(uncompressed, ulength); + return RawUncompress(compressed, n, string_as_array(uncompressed)); +} + +// A Writer that drops everything on the floor and just does validation +class SnappyDecompressionValidator { + private: + size_t expected_; + size_t produced_; + + public: + inline SnappyDecompressionValidator() : expected_(0), produced_(0) { } + inline void SetExpectedLength(size_t len) { + expected_ = len; + } + inline bool CheckLength() const { + return expected_ == produced_; + } + inline bool Append(const char* ip, size_t len) { + produced_ += len; + return produced_ <= expected_; + } + inline bool TryFastAppend(const char* ip, size_t available, size_t length) { + return false; + } + inline bool AppendFromSelf(size_t offset, size_t len) { + // See SnappyArrayWriter::AppendFromSelf for an explanation of + // the "offset - 1u" trick. + if (produced_ <= offset - 1u) return false; + produced_ += len; + return produced_ <= expected_; + } + inline void Flush() {} +}; + +bool IsValidCompressedBuffer(const char* compressed, size_t n) { + ByteArraySource reader(compressed, n); + SnappyDecompressionValidator writer; + return InternalUncompress(&reader, &writer); +} + +bool IsValidCompressed(Source* compressed) { + SnappyDecompressionValidator writer; + return InternalUncompress(compressed, &writer); +} + +void RawCompress(const char* input, + size_t input_length, + char* compressed, + size_t* compressed_length) { + ByteArraySource reader(input, input_length); + UncheckedByteArraySink writer(compressed); + Compress(&reader, &writer); + + // Compute how many bytes were added + *compressed_length = (writer.CurrentDestination() - compressed); +} + +size_t Compress(const char* input, size_t input_length, string* compressed) { + // Pre-grow the buffer to the max length of the compressed output + compressed->resize(MaxCompressedLength(input_length)); + + size_t compressed_length; + RawCompress(input, input_length, string_as_array(compressed), + &compressed_length); + compressed->resize(compressed_length); + return compressed_length; +} + +// ----------------------------------------------------------------------- +// Sink interface +// ----------------------------------------------------------------------- + +// A type that decompresses into a Sink. The template parameter +// Allocator must export one method "char* Allocate(int size);", which +// allocates a buffer of "size" and appends that to the destination. +template +class SnappyScatteredWriter { + Allocator allocator_; + + // We need random access into the data generated so far. Therefore + // we keep track of all of the generated data as an array of blocks. + // All of the blocks except the last have length kBlockSize. + vector blocks_; + size_t expected_; + + // Total size of all fully generated blocks so far + size_t full_size_; + + // Pointer into current output block + char* op_base_; // Base of output block + char* op_ptr_; // Pointer to next unfilled byte in block + char* op_limit_; // Pointer just past block + + inline size_t Size() const { + return full_size_ + (op_ptr_ - op_base_); + } + + bool SlowAppend(const char* ip, size_t len); + bool SlowAppendFromSelf(size_t offset, size_t len); + + public: + inline explicit SnappyScatteredWriter(const Allocator& allocator) + : allocator_(allocator), + full_size_(0), + op_base_(NULL), + op_ptr_(NULL), + op_limit_(NULL) { + } + + inline void SetExpectedLength(size_t len) { + assert(blocks_.empty()); + expected_ = len; + } + + inline bool CheckLength() const { + return Size() == expected_; + } + + // Return the number of bytes actually uncompressed so far + inline size_t Produced() const { + return Size(); + } + + inline bool Append(const char* ip, size_t len) { + size_t avail = op_limit_ - op_ptr_; + if (len <= avail) { + // Fast path + memcpy(op_ptr_, ip, len); + op_ptr_ += len; + return true; + } else { + return SlowAppend(ip, len); + } + } + + inline bool TryFastAppend(const char* ip, size_t available, size_t length) { + char* op = op_ptr_; + const int space_left = op_limit_ - op; + if (length <= 16 && available >= 16 + kMaximumTagLength && + space_left >= 16) { + // Fast path, used for the majority (about 95%) of invocations. + UNALIGNED_STORE64(op, UNALIGNED_LOAD64(ip)); + UNALIGNED_STORE64(op + 8, UNALIGNED_LOAD64(ip + 8)); + op_ptr_ = op + length; + return true; + } else { + return false; + } + } + + inline bool AppendFromSelf(size_t offset, size_t len) { + // See SnappyArrayWriter::AppendFromSelf for an explanation of + // the "offset - 1u" trick. + if (offset - 1u < op_ptr_ - op_base_) { + const size_t space_left = op_limit_ - op_ptr_; + if (space_left >= len + kMaxIncrementCopyOverflow) { + // Fast path: src and dst in current block. + IncrementalCopyFastPath(op_ptr_ - offset, op_ptr_, len); + op_ptr_ += len; + return true; + } + } + return SlowAppendFromSelf(offset, len); + } + + // Called at the end of the decompress. We ask the allocator + // write all blocks to the sink. + inline void Flush() { allocator_.Flush(Produced()); } +}; + +template +bool SnappyScatteredWriter::SlowAppend(const char* ip, size_t len) { + size_t avail = op_limit_ - op_ptr_; + while (len > avail) { + // Completely fill this block + memcpy(op_ptr_, ip, avail); + op_ptr_ += avail; + assert(op_limit_ - op_ptr_ == 0); + full_size_ += (op_ptr_ - op_base_); + len -= avail; + ip += avail; + + // Bounds check + if (full_size_ + len > expected_) { + return false; + } + + // Make new block + size_t bsize = min(kBlockSize, expected_ - full_size_); + op_base_ = allocator_.Allocate(bsize); + op_ptr_ = op_base_; + op_limit_ = op_base_ + bsize; + blocks_.push_back(op_base_); + avail = bsize; + } + + memcpy(op_ptr_, ip, len); + op_ptr_ += len; + return true; +} + +template +bool SnappyScatteredWriter::SlowAppendFromSelf(size_t offset, + size_t len) { + // Overflow check + // See SnappyArrayWriter::AppendFromSelf for an explanation of + // the "offset - 1u" trick. + const size_t cur = Size(); + if (offset - 1u >= cur) return false; + if (expected_ - cur < len) return false; + + // Currently we shouldn't ever hit this path because Compress() chops the + // input into blocks and does not create cross-block copies. However, it is + // nice if we do not rely on that, since we can get better compression if we + // allow cross-block copies and thus might want to change the compressor in + // the future. + size_t src = cur - offset; + while (len-- > 0) { + char c = blocks_[src >> kBlockLog][src & (kBlockSize-1)]; + Append(&c, 1); + src++; + } + return true; +} + +class SnappySinkAllocator { + public: + explicit SnappySinkAllocator(Sink* dest): dest_(dest) {} + ~SnappySinkAllocator() {} + + char* Allocate(int size) { + Datablock block(new char[size], size); + blocks_.push_back(block); + return block.data; + } + + // We flush only at the end, because the writer wants + // random access to the blocks and once we hand the + // block over to the sink, we can't access it anymore. + // Also we don't write more than has been actually written + // to the blocks. + void Flush(size_t size) { + size_t size_written = 0; + size_t block_size; + for (int i = 0; i < blocks_.size(); ++i) { + block_size = min(blocks_[i].size, size - size_written); + dest_->AppendAndTakeOwnership(blocks_[i].data, block_size, + &SnappySinkAllocator::Deleter, NULL); + size_written += block_size; + } + blocks_.clear(); + } + + private: + struct Datablock { + char* data; + size_t size; + Datablock(char* p, size_t s) : data(p), size(s) {} + }; + + static void Deleter(void* arg, const char* bytes, size_t size) { + delete[] bytes; + } + + Sink* dest_; + vector blocks_; + + // Note: copying this object is allowed +}; + +size_t UncompressAsMuchAsPossible(Source* compressed, Sink* uncompressed) { + SnappySinkAllocator allocator(uncompressed); + SnappyScatteredWriter writer(allocator); + InternalUncompress(compressed, &writer); + return writer.Produced(); +} + +bool Uncompress(Source* compressed, Sink* uncompressed) { + // Read the uncompressed length from the front of the compressed input + SnappyDecompressor decompressor(compressed); + uint32 uncompressed_len = 0; + if (!decompressor.ReadUncompressedLength(&uncompressed_len)) { + return false; + } + + char c; + size_t allocated_size; + char* buf = uncompressed->GetAppendBufferVariable( + 1, uncompressed_len, &c, 1, &allocated_size); + + // If we can get a flat buffer, then use it, otherwise do block by block + // uncompression + if (allocated_size >= uncompressed_len) { + SnappyArrayWriter writer(buf); + bool result = InternalUncompressAllTags( + &decompressor, &writer, uncompressed_len); + uncompressed->Append(buf, writer.Produced()); + return result; + } else { + SnappySinkAllocator allocator(uncompressed); + SnappyScatteredWriter writer(allocator); + return InternalUncompressAllTags(&decompressor, &writer, uncompressed_len); + } +} + +} // end namespace snappy