LibXDiff

Section: File Differential Library (3)
Updated: 0.23
Index Return to Main Contents
 

NAME

xdl_set_allocator, xdl_malloc, xdl_free, xdl_realloc, xdl_init_mmfile, xdl_free_mmfile, xdl_mmfile_iscompact, xdl_seek_mmfile, xdl_read_mmfile, xdl_write_mmfile, xdl_writem_mmfile, xdl_mmfile_writeallocate, xdl_mmfile_ptradd, xdl_mmfile_first, xdl_mmfile_next, xdl_mmfile_size, xdl_mmfile_cmp, xdl_mmfile_compact, xdl_diff, xdl_patch, xdl_merge3, xdl_bdiff_mb, xdl_bdiff, xdl_rabdiff_mb, xdl_rabdiff, xdl_bdiff_tgsize, xdl_bpatch - File Differential Library support functions

 

SYNOPSIS

#include <xdiff.h>

int xdl_set_allocator(memallocator_t const *malt);
void *xdl_malloc(unsigned int size);
void xdl_free(void *ptr);
void *xdl_realloc(void *ptr, unsigned int nsize);
int xdl_init_mmfile(mmfile_t *mmf, long bsize, unsigned long flags);
void xdl_free_mmfile(mmfile_t *mmf);
int xdl_mmfile_iscompact(mmfile_t *mmf);
int xdl_seek_mmfile(mmfile_t *mmf, long off);
long xdl_read_mmfile(mmfile_t *mmf, void *data, long size);
long xdl_write_mmfile(mmfile_t *mmf, void const *data, long size);
long xdl_writem_mmfile(mmfile_t *mmf, mmbuffer_t *mb, int nbuf);
void *xdl_mmfile_writeallocate(mmfile_t *mmf, long size);
long xdl_mmfile_ptradd(mmfile_t *mmf, char *ptr, long size, unsigned long flags);
void *xdl_mmfile_first(mmfile_t *mmf, long *size);
void *xdl_mmfile_next(mmfile_t *mmf, long *size);
long xdl_mmfile_size(mmfile_t *mmf);
int xdl_mmfile_cmp(mmfile_t *mmf1, mmfile_t *mmf2);
int xdl_mmfile_compact(mmfile_t *mmfo, mmfile_t *mmfc, long bsize, unsigned long flags);
int xdl_diff(mmfile_t *mmf1, mmfile_t *mmf2, xpparam_t const *xpp, xdemitconf_t const *xecfg, xdemitcb_t *ecb);
int xdl_patch(mmfile_t *mmf, mmfile_t *mmfp, int mode, xdemitcb_t *ecb, xdemitcb_t *rjecb);
int xdl_merge3(mmfile_t *mmfo, mmfile_t *mmf1, mmfile_t *mmf2, xdemitcb_t *ecb, xdemitcb_t *rjecb);
int xdl_bdiff_mb(mmbuffer_t *mmb1, mmbuffer_t *mmb2, bdiffparam_t const *bdp, xdemitcb_t *ecb);
int xdl_bdiff(mmfile_t *mmf1, mmfile_t *mmf2, bdiffparam_t const *bdp, xdemitcb_t *ecb);
int xdl_rabdiff_mb(mmbuffer_t *mmb1, mmbuffer_t *mmb2, xdemitcb_t *ecb);
int xdl_rabdiff(mmfile_t *mmf1, mmfile_t *mmf2, xdemitcb_t *ecb);
long xdl_bdiff_tgsize(mmfile_t *mmfp);
int xdl_bpatch(mmfile_t *mmf, mmfile_t *mmfp, xdemitcb_t *ecb);

 

DESCRIPTION

The LibXDiff library implements basic and yet complete functionalities to create file differences/patches to both binary and text files. The library uses memory files as file abstraction to achieve both performance and portability. For binary files, LibXDiff implements both (with some modification) the algorithm described in File System Support for Delta Compression by Joshua P. MacDonald, and the method described in Fingerprinting by Random Polynomials by Michael O. Rabin. While for text files it follows directives described in An O(ND) Difference Algorithm and Its Variations by Eugene W. Myers. Memory files used by the library are basically a collection of buffers that store the file content. There are two different requirements for memory files when passed to diff/patch functions. Text files for diff/patch functions require that a single line do not have to spawn across two different memory file blocks. Binary diff/patch functions require memory files to be compact. A compact memory files is a file whose content is stored inside a single block. Functionalities inside the library are available to satisfy these rules. Using the XDL_MMF_ATOMIC memory file flag it is possible to make writes to not split the written record across different blocks, while the functions xdl_mmfile_iscompact() , xdl_mmfile_compact() and xdl_mmfile_writeallocate() are usefull to test if the file is compact and to create a compacted version of the file itself. The text file differential output uses the raw unified output format, by omitting the file header since the result is always relative to a single compare operation (between two files). The output format of the binary patch file is proprietary (and binary) and it is basically a collection of copy and insert commands, like described inside the MacDonald paper.

 

Functions

The following functions are defined:
int xdl_set_allocator(memallocator_t const *malt);

The LibXDiff library enable the user to set its own memory allocator, that will be used for all the following memory requests. The allocator must be set before to start calling the LibXDiff library with a call to xdl_set_allocator(). The memory allocator structure contains the following members:


        typedef struct s_memallocator {
                void *priv;
                void *(*malloc)(void *priv, unsigned int size);
                void (*free)(void *priv, void *ptr);
                void *(*realloc)(void *priv, void *ptr, unsigned int nsize);
        } memallocator_t;

The malloc() function pointer will be used by LibXDiff to request a memory block of size bytes. The free() function pointer will be called to free a previously allocated block ptr , while the realloc() will be used to resize the ptr to a new nsize size in bytes. The priv structure member will be passed to the malloc(),free(),realloc() functions as first parameter. The LibXDiff user must call xdl_set_allocator() before starting using the library, otherwise LibXDiff functions will fail due to the lack of memory allocation support. A typical initialization sequence for POSIX systems will use the standard malloc(3), free(3), realloc(3) and will look like:

        void *wrap_malloc(void *priv, unsigned int size) {
                return malloc(size);
        }

        void wrap_free(void *priv, void *ptr) {
                free(ptr);
        }

        void *wrap_realloc(void *priv, void *ptr, unsigned int size) {
                return realloc(ptr, size);
        }

        void my_init_xdiff(void) {
                memallocator_t malt;

                malt.priv = NULL;
                malt.malloc = wrap_malloc;
                malt.free = wrap_free;
                malt.realloc = wrap_realloc;
                xdl_set_allocator(&malt);
        }

void *xdl_malloc(unsigned int size);

Allocates a memory block of size bytes using the LibXDiff memory allocator. The user can specify its own allocator using the xdl_set_allocator() function. The xdl_malloc() return a pointer to the newly allocated block, or NULL in case of failure.

void xdl_free(void *ptr);

Free a previously allocated memory block pointed by ptr. The ptr block must has been allocated using either xdl_malloc() or xdl_realloc().

void *xdl_realloc(void *ptr, unsigned int nsize);

Resizes the memory block pointed by ptr to a new size nsize. Return the resized block if successful, or NULL in case the reallocation fails. After a successful reallocation, the old ptr block is to be considered no more valid.

int xdl_init_mmfile(mmfile_t *mmf, long bsize, unsigned long flags);

Initialize the memory file mmf by requiring an internal block size of bsize. The flags parameter is a combination of the following flags :

XDL_MMF_ATOMIC Writes on the memory file will be atomic. That is, the data will not be split on two or more different blocks.

Once an xdl_init_mmfile() succeeded, a matching xdl_free_mmfile() must be called when the user has done using the memory file, otherwise serious memory leaks will happen. The function return 0 if succeed or -1 if an error is encountered.

void xdl_free_mmfile(mmfile_t *mmf);

Free all the data associated with the mmf memory file.

int xdl_mmfile_iscompact(mmfile_t *mmf);

Returns an integer different from 0 if the mmf memory file is compact, 0 otherwise. A compact memory file is one that have the whole content stored inside a single block.

int xdl_seek_mmfile(mmfile_t *mmf, long off);

Set the current data pointer of the memory file mmf to the specified offset off from the beginning of the file itself. Returns 0 if successful or -1 if an error happened.

long xdl_read_mmfile(mmfile_t *mmf, void *data, long size);

Request to read size bytes from the memory file mmf by storing the data inside the data buffer. Returns the number of bytes read into the data buffer. The amount of data read can be lower than the specified size. The function returns -1 if an error happened.

long xdl_write_mmfile(mmfile_t *mmf, void const *data, long size);

Request to write size bytes from the specified buffer data into the memory file mmf. If the memory file has been created using the XDL_MMF_ATOMIC flag, the write request will not be split across different blocks. Note that all write operations done on memory files do append data at the end the file, and writes in the middle of it are allowed. This is because the library memory file abstraction does not need this functionality to be available. The function returns the number of bytes written or a number lower than size if an error happened.

long xdl_writem_mmfile(mmfile_t *mmf, mmbuffer_t *mb, int nbuf);

Request to sequentially write nbuf memory buffers passed inside the array mb into the memory file mmf. The memory buffer structure is defined as :


        typedef struct s_mmbuffer {
                char *ptr;
                long size;
        } mmbuffer_t;

The ptr field is a pointer to the user data, whose size is specified inside the size structure field. The function returns the total number of bytes written or a lower number if an error happened.

void *xdl_mmfile_writeallocate(mmfile_t *mmf, long size);

The function request to allocate a write buffer of size bytes in the mmf memory file and returns the pointer to the allocated buffer. The user will have the responsibility to store size bytes (no more, no less) inside the memory region pointed to by the returned pointer. The files size will grow of size bytes as a consequence of this operation. The function will return NULL if an error happened.

long xdl_mmfile_ptradd(mmfile_t *mmf, char *ptr, long size, unsigned long flags);

The function adds a user specified block to the end of the memory file mmf. The block first byte is pointed to by ptr and its length is size bytes. The flags parameter can be used to specify attributes of the user memory block. Currently supported attributes are:

XDL_MMB_READONLY Specify that the added memory block must be treated as read-only, and every attempt to write on it should result in a failure of the memory file writing functions.

The purpose of this function is basically to avoid copying memory around, by helping the library to not drain the CPU cache. The function returns size in case of success, or -1 in case of error.

void *xdl_mmfile_first(mmfile_t *mmf, long *size);

The function is used to return the first block of the mmf memory file block chain. The size parameter will receive the size of the block, while the function will return the pointer the the first byte of the block itself. The function returns NULL if the file is empty.

void *xdl_mmfile_next(mmfile_t *mmf, long *size);

The function is used to return the next block of the mmf memory file block chain. The size parameter will receive the size of the block, while the function will return the pointer the the first byte of the block itself. The function returns NULL if the current block is the last one of the chain.

long xdl_mmfile_size(mmfile_t *mmf);

The function returns the size of the specified memory file mmf.

int xdl_mmfile_cmp(mmfile_t *mmf1, mmfile_t *mmf2);

Request to compare two memory files mmf1 and mmf2 and returns 0 if files are identical, or a value different from 0 if files are different.

int xdl_mmfile_compact(mmfile_t *mmfo, mmfile_t *mmfc, long bsize, unsigned long flags);

Request to create a compact version of the memory file mmfo into the (uninitialized) memory file mmfc. The bsize parameter specify the requested block size and flags specify flags to be used to create the new mmfc memory file (see xdl_init_mmfile() ). The function returns 0 if succedded or -1 if an error happened.

int xdl_diff(mmfile_t *mmf1, mmfile_t *mmf2, xpparam_t const *xpp, xdemitconf_t const *xecfg, xdemitcb_t *ecb);

Request to create the difference between the two text memory files mmf1 and mmf2. The mmf1 memory files is considered the "old" file while mmf2 is considered the "new" file. So the function will create a patch file that once applied to mmf1 will give mmf2 as result. Files mmf1 and mmf2 must be atomic from a line point of view (or, as an extreme, compact), that means that a single test line cannot spread among different memory file blocks. The xpp parameter is a pointer to a structure :


        typedef struct s_xpparam {
                unsigned long flags;
        } xpparam_t;

that is used to specify parameters to be used by the file differential algorithm. The flags field is a combination of the following flags :

XDF_NEED_MINIMAL Requires the minimal edit script to be found by the algorithm (may be slow).

The xecfg parameter point to a structure :


        typedef struct s_xdemitconf {
                long ctxlen;
        } xdemitconf_t;

that is used to configure the algorithm responsible of the creation the the differential file from an edit script. The ctxlen field is used to specify the amount of context to be emitted inside the differential file (the value 3 is suggested for normal operations). The parameter ecb is a pointer to a structure :

        typedef struct s_xdemitcb {
                void *priv;
                int (*outf)(void *, mmbuffer_t *, int);
        } xdemitcb_t;

that is used by the differential file creation algorithm to emit the created data. The priv field is an opaque pointer to a user specified data, while the outf field point to a callback function that is called internally to emit algorithm generated data rappresenting the differential file. The first parameter of the callback is the same priv field specified inside the xdemitcb_t structure. The second parameter point to an array of mmbuffer_t (see above for a definition of the structure) whose element count is specified inside the last parameter of the callback itself. The callback will always be called with entire records (lines) and never a record (line) will be emitted using two different callback calls. This is important because if the called will use another memory file to store the result, by creating the target memory file with XDL_MMF_ATOMIC will guarantee the "atomicity" of the memory file itself. The function returns 0 if succeeded or -1 if an error occurred.

int xdl_patch(mmfile_t *mmf, mmfile_t *mmfp, int mode, xdemitcb_t *ecb, xdemitcb_t *rjecb);

Request to patch the memory file mmf using the patch file stored in mmfp. The mmf memory file is not changed during the operation and can be considered as read only. The mode parameter can be one of the following values :

XDL_PATCH_NORMAL Perform standard patching like if the patch memory file mmfp has been created using mmf as "old" file.
XDL_PATCH_REVERSE Apply the reverse patch. That means that the mmf memory file has to be considered as if it was specified as "new" file during the differential operation ( xdl_diff() ). The result of the operation will then be the file content that was used as "old" file during the differential operation.

The following flags can be specified (by or-ing them) to one of the above:

XDL_PATCH_IGNOREBSPACE Ignore the whitespace at the beginning and the end of the line.

The ecb will be used by the patch algorithm to create the result file while the rjecb will be used to emit all differential chunks that cannot be applied. Like explained above, callbacks are always called with entire records to guarantee atomicity of the resulting output. The function returns 0 if succeeded without performing any fuzzy hunk detection, a positive value if it secceeded with fuzzy hunk detection or -1 if an error occurred during the patch operation.

int xdl_merge3(mmfile_t *mmfo, mmfile_t *mmf1, mmfile_t *mmf2, xdemitcb_t *ecb, xdemitcb_t *rjecb);

Merges three files together. The mmfo file is the original one, while mmf1 and mmf2 are two modified versions of mmfo. The function works by creating a differential between mmfo and mmf2 and by applying the resulting patch to mmf1. Because of this sequence, mmf1 changes will be privileged against the ones of mmf2. The ecb will be used by the patch algorithm to create the result file while the rjecb will be used to emit all differential chunks that cannot be applied. Like explained above, callbacks are always called with entire records to guarantee atomicity of the resulting output. The function returns 0 if succeeded or -1 if an error occurred during the patch operation.

int xdl_bdiff(mmfile_t *mmf1, mmfile_t *mmf2, bdiffparam_t const *bdp, xdemitcb_t *ecb);

Request to create the difference between the two text memory files mmf1 and mmf2. The mmf1 memory files is considered the "old" file while mmf2 is considered the "new" file. So the function will create a patch file that once applied to mmf1 will give mmf2 as result. Files mmf1 and mmf2 must be compact to make it easy and faster to perform the difference operation. Functions are available to check for compactness ( xdl_mmfile_iscompact() ) and to make compact a non-compact file ( xdl_mmfile_compact() ). An example of how to create a compact memory file (described inside the test subdirectory) is :


        int xdlt_load_mmfile(char const *fname, mmfile_t *mf, int binmode) {
                char cc;
                int fd;
                long size, bsize;
                char *blk;

                if (xdl_init_mmfile(mf, XDLT_STD_BLKSIZE, XDL_MMF_ATOMIC) < 0)
                        return -1;
                if ((fd = open(fname, O_RDONLY)) == -1) {
                        perror(fname);
                        xdl_free_mmfile(mf);
                        return -1;
                }
                if ((size = bsize = lseek(fd, 0, SEEK_END)) > 0 && !binmode) {
                        if (lseek(fd, -1, SEEK_END) != (off_t) -1 &&
                            read(fd, &cc, 1) && cc != '\n')
                                bsize++;
                }
                lseek(fd, 0, SEEK_SET);
                if (!(blk = (char *) xdl_mmfile_writeallocate(mf, bsize))) {
                        xdl_free_mmfile(mf);
                        close(fd);
                        return -1;
                }
                if (read(fd, blk, (size_t) size) != (size_t) size) {
                        perror(fname);
                        xdl_free_mmfile(mf);
                        close(fd);
                        return -1;
                }
                close(fd);
                if (bsize > size)
                        blk[size] = '\n';
                return 0;
        }

The bdp parameter points to a structure :

        typedef struct s_bdiffparam {
                long bsize;
        } bdiffparam_t;

that is used to pass information to the binary file differential algorithm. The bsize parameter specify the size of the block that will be used to decompose mmf1 during the block classification phase of the algorithm (see MacDonald paper). Suggested values go from 16 to 64, with a preferred power of two characteristic. The ecb parameter is used to pass the emission callback to the algorithm responsible of the output file creation. The function returns 0 if succeede or -1 if an error is occurred.

int xdl_bdiff_mb(mmbuffer_t *mmb1, mmbuffer_t *mmb2, bdiffparam_t const *bdp, xdemitcb_t *ecb);

Same as xdl_bdiff() but it works on memory buffer directly. The xdl_bdiff() is implemented internally with a xdl_bdiff_mb() after having setup the two memory buffers from the passed memory files (that must be compact, as described above). The memory buffer structure is defined as :


        typedef struct s_mmbuffer {
                char *ptr;
                long size;
        } mmbuffer_t;

An empty memory buffer is specified by setting the ptr member as NULL and the size member as zero. The reason of having this function is to avoid the memory file preparation, that might involve copying memory from other sources. Using the xdl_bdiff_mb(), the caller can setup the two memory buffer by using, for example, mmap(2), and hence avoiding unnecessary memory copies. The other parameters and the return value of the function xdl_bdiff_mb() are the same as the ones already described in xdl_bdiff().

int xdl_rabdiff(mmfile_t *mmf1, mmfile_t *mmf2, xdemitcb_t *ecb);

Request to create the difference between the two text memory files mmf1 and mmf2 using the Rabin's polynomial fingerprinting algorithm. This algorithm typically performs faster and produces smaller deltas, when compared to the XDelta-like one. The mmf1 memory files is considered the "old" file while mmf2 is considered the "new" file. So the function will create a patch file that once applied to mmf1 will give mmf2 as result. Files mmf1 and mmf2 must be compact to make it easy and faster to perform the difference operation. Functions are available to check for compactness ( xdl_mmfile_iscompact() ) and to make compact a non-compact file ( xdl_mmfile_compact() ). The ecb parameter is used to pass the emission callback to the algorithm responsible of the output file creation. The function returns 0 if succeede or -1 if an error is occurred.

int xdl_rabdiff_mb(mmbuffer_t *mmb1, mmbuffer_t *mmb2, xdemitcb_t *ecb);

Same as xdl_rabdiff() but it works on memory buffer directly. The memory buffer structure is defined as :


        typedef struct s_mmbuffer {
                char *ptr;
                long size;
        } mmbuffer_t;

An empty memory buffer is specified by setting the ptr member as NULL and the size member as zero. The reason of having this function is to avoid the memory file preparation, that might involve copying memory from other sources. Using the xdl_rabdiff_mb(), the caller can setup the two memory buffer by using, for example, mmap(2), and hence avoiding unnecessary memory copies. The other parameters and the return value of the function xdl_rabdiff_mb() are the same as the ones already described in xdl_rabdiff().

long xdl_bdiff_tgsize(mmfile_t *mmfp);

Given a binary memory file patch, it returns the size that the result file will have once the patch is applied to the target file. It can be used to pre-allocate (or write-allocate) a memory block to store the patch result so that a compact file will be available at the end of the operation. The function returns the requested size, or -1 if an error occurred during the operation.

int xdl_bpatch(mmfile_t *mmf, mmfile_t *mmfp, xdemitcb_t *ecb);

Request to patch the binary memory file mmf using the binary patch file stored in mmfp. The mmf memory file is not changed during the operation and can be considered as read only. The binary patch algorithm has no notion of context, so the patch operation cannot be partial (either success or failure). The ecb parameter contain the callabck (see above for description) used by the binary patch algorithm to emit the result file. The function returns 0 if succeeded or -1 if an error occurred during the patch operation.

 

SEE ALSO

Two papers drove the content of this library and these are :
o
File System Support for Delta Compression by Joshua P. MacDonald http://www.xmailserver.org/xdfs.pdf
o
Fingerprinting by Random Polynomials by Michael O. Rabin http://www.xmailserver.org/rabin.pdf
o
An O(ND) Difference Algorithm and Its Variations by Eugene W. Myers http://www.xmailserver.org/diff2.pdf

Also usefull information can be looked up inside the diffutil GNU package :

http://www.gnu.org/software/diffutils/diffutils.html

 

LICENSE

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. A copy of the license is available at :

http://www.gnu.org/copyleft/lesser.html

 

AUTHOR

Developed by Davide Libenzi <davidel@xmailserver.org>

 

AVAILABILITY

The latest version of LibXDiff can be found at :

http://www.xmailserver.org/xdiff-lib.html

 

BUGS

There are no known bugs. Bug reports and comments to Davide Libenzi <davidel@xmailserver.org>


 

Index

NAME
SYNOPSIS
DESCRIPTION
Functions
SEE ALSO
LICENSE
AUTHOR
AVAILABILITY
BUGS

This document was created by man2html, using the manual pages.
Time: 21:26:59 GMT, November 12, 2008