Commit 44b1ff4c by Vicent Martí

filter: Apply filters before writing a file to the ODB

Initial implementation. The relevant code is in `blob.c`: the blob write
function has been split into smaller functions.

	- Directly write a file to the ODB in streaming mode

	- Directly write a symlink to the ODB in direct mode

	- Apply a filter, and write a file to the ODB in direct mode

When trying to write a file, we first call `git_filter__load_for_file`,
which populates a filters array with the required filters based on the
filename.

If no filters are resolved to the filename, we can write to the ODB in
streaming mode straight from disk. Otherwise, we load the whole file in
memory and use double-buffering to apply the filter chain. We finish
by writing the file as a whole to the ODB.
parent 13224ea4
......@@ -11,6 +11,7 @@
#include "common.h"
#include "blob.h"
#include "filter.h"
const void *git_blob_rawcontent(git_blob *blob)
{
......@@ -65,15 +66,101 @@ int git_blob_create_frombuffer(git_oid *oid, git_repository *repo, const void *b
return GIT_SUCCESS;
}
static int write_file_stream(git_oid *oid, git_odb *odb, const char *path, git_off_t file_size)
{
int fd, error;
char buffer[4096];
git_odb_stream *stream = NULL;
if ((error = git_odb_open_wstream(&stream, odb, file_size, GIT_OBJ_BLOB)) < GIT_SUCCESS)
return error;
if ((fd = p_open(path, O_RDONLY)) < 0) {
error = git__throw(GIT_ENOTFOUND, "Failed to create blob. Could not open '%s'", path);
goto cleanup;
}
while (file_size > 0) {
ssize_t read_len = p_read(fd, buffer, sizeof(buffer));
if (read_len < 0) {
error = git__throw(GIT_EOSERR, "Failed to create blob. Can't read full file");
p_close(fd);
goto cleanup;
}
stream->write(stream, buffer, read_len);
file_size -= read_len;
}
p_close(fd);
error = stream->finalize_write(oid, stream);
cleanup:
stream->free(stream);
return error;
}
static int write_file_filtered(
git_oid *oid,
git_odb *odb,
const char *path,
git_vector *filters)
{
int error;
git_buf file_in = GIT_BUF_INIT;
git_buf filter_result = GIT_BUF_INIT;
error = git_futils_readbuffer(&file_in, path);
if (error < GIT_SUCCESS)
return error;
error = git_filter__apply(&filter_result, &file_in, filters, path);
if (error < GIT_SUCCESS) {
git_buf_free(&file_in);
git_buf_free(&filter_result);
return error;
}
error = git_odb_write(oid, odb, filter_result.ptr, filter_result.size, GIT_OBJ_BLOB);
git_buf_free(&file_in);
git_buf_free(&filter_result);
return GIT_SUCCESS;
}
static int write_symlink(git_oid *oid, git_odb *odb, const char *path, size_t link_size)
{
char *link_data;
ssize_t read_len;
int error;
link_data = git__malloc(link_size);
if (!link_data)
return GIT_ENOMEM;
read_len = p_readlink(path, link_data, link_size);
if (read_len != (ssize_t)link_size) {
free(link_data);
return git__throw(GIT_EOSERR, "Failed to create blob. Can't read symlink");
}
error = git_odb_write(oid, odb, (void *)link_data, link_size, GIT_OBJ_BLOB);
free(link_data);
return error;
}
int git_blob_create_fromfile(git_oid *oid, git_repository *repo, const char *path)
{
int error = GIT_SUCCESS;
git_buf full_path = GIT_BUF_INIT;
git_off_t size;
git_odb_stream *stream = NULL;
struct stat st;
const char *workdir;
git_odb *odb;
git_odb *odb = NULL;
workdir = git_repository_workdir(repo);
if (workdir == NULL)
......@@ -95,63 +182,36 @@ int git_blob_create_fromfile(git_oid *oid, git_repository *repo, const char *pat
if (error < GIT_SUCCESS)
goto cleanup;
if ((error = git_odb_open_wstream(&stream, odb, (size_t)size, GIT_OBJ_BLOB)) < GIT_SUCCESS)
goto cleanup;
if (S_ISLNK(st.st_mode)) {
char *link_data;
ssize_t read_len;
link_data = git__malloc((size_t)size);
if (!link_data) {
error = GIT_ENOMEM;
goto cleanup;
}
read_len = p_readlink(full_path.ptr, link_data, (size_t)size);
if (read_len != (ssize_t)size) {
error = git__throw(GIT_EOSERR, "Failed to create blob. Can't read symlink");
free(link_data);
goto cleanup;
}
stream->write(stream, link_data, (size_t)size);
free(link_data);
error = write_symlink(oid, odb, full_path.ptr, (size_t)size);
} else {
int fd;
char buffer[2048];
if ((fd = p_open(full_path.ptr, O_RDONLY)) < 0) {
error = git__throw(GIT_ENOTFOUND, "Failed to create blob. Could not open '%s'", full_path.ptr);
goto cleanup;
}
git_vector write_filters = GIT_VECTOR_INIT;
while (size > 0) {
ssize_t read_len = p_read(fd, buffer, sizeof(buffer));
if (read_len < 0) {
error = git__throw(GIT_EOSERR, "Failed to create blob. Can't read full file");
p_close(fd);
if ((error = git_filter__load_for_file(
&write_filters, repo, full_path.ptr, GIT_FILTER_TO_ODB)) < GIT_SUCCESS)
goto cleanup;
}
stream->write(stream, buffer, read_len);
size -= read_len;
if (write_filters.length == 0) {
error = write_file_stream(oid, odb, full_path.ptr, size);
} else {
error = write_file_filtered(oid, odb, full_path.ptr, &write_filters);
}
p_close(fd);
/*
* TODO: eventually support streaming filtered files, for files which are bigger
* than a given threshold. This is not a priority because applying a filter in
* streaming mode changes the final size of the blob, and without knowing its
* final size, the blob cannot be written in stream mode to the ODB.
*
* The plan is to do streaming writes to a tempfile on disk and then opening
* streaming that file to the ODB, using `write_file_stream`.
*
* CAREFULLY DESIGNED APIS YO
*/
}
error = stream->finalize_write(oid, stream);
cleanup:
if (stream)
stream->free(stream);
git_buf_free(&full_path);
return error;
}
......@@ -235,7 +235,7 @@ char *git_buf_detach(git_buf *buf)
{
char *data = buf->ptr;
if (buf->asize <= 0)
if (buf->asize == 0 || buf->ptr == &git_buf__oom)
return NULL;
git_buf_init(buf, 0);
......
/*
* Copyright (C) 2009-2012 the libgit2 contributors
*
* This file is part of libgit2, distributed under the GNU GPL v2 with
* a Linking Exception. For full terms see the included COPYING file.
*/
#include "common.h"
#include "fileops.h"
#include "hash.h"
#include "filter.h"
/* Fresh from Core Git. I wonder what we could use this for... */
void git_text__stat(git_text_stats *stats, git_buf *text)
{
size_t i;
memset(stats, 0, sizeof(*stats));
for (i = 0; i < text->size; i++) {
unsigned char c = text->ptr[i];
if (c == '\r') {
stats->cr++;
if (i + 1 < text->size && text->ptr[i + 1] == '\n')
stats->crlf++;
continue;
}
if (c == '\n') {
stats->lf++;
continue;
}
if (c == 127)
/* DEL */
stats->nonprintable++;
else if (c < 32) {
switch (c) {
/* BS, HT, ESC and FF */
case '\b': case '\t': case '\033': case '\014':
stats->printable++;
break;
case 0:
stats->nul++;
/* fall through */
default:
stats->nonprintable++;
}
}
else
stats->printable++;
}
/* If file ends with EOF then don't count this EOF as non-printable. */
if (text->size >= 1 && text->ptr[text->size - 1] == '\032')
stats->nonprintable--;
}
/*
* Fresh from Core Git
*/
int git_text__is_binary(git_text_stats *stats)
{
if (stats->nul)
return 1;
if ((stats->printable >> 7) < stats->nonprintable)
return 1;
/*
* Other heuristics? Average line length might be relevant,
* as might LF vs CR vs CRLF counts..
*
* NOTE! It might be normal to have a low ratio of CRLF to LF
* (somebody starts with a LF-only file and edits it with an editor
* that adds CRLF only to lines that are added..). But do we
* want to support CR-only? Probably not.
*/
return 0;
}
int git_filter__load_for_file(git_vector *filters, git_repository *repo, const char *full_path, int mode)
{
/* We don't load any filters yet. HAHA */
return 0;
}
int git_filter__apply(git_buf *dest, git_buf *source, git_vector *filters, const char *filename)
{
unsigned int src, dst, i;
git_buf *dbuffer[2];
dbuffer[0] = source;
dbuffer[1] = dest;
src = 0;
/* Pre-grow the destination buffer to more or less the size
* we expect it to have */
if (git_buf_grow(dest, source->size) < 0)
return GIT_ENOMEM;
for (i = 0; i < filters->length; ++i) {
git_filter_cb filter = git_vector_get(filters, i);
dst = (src + 1) % 2;
git_buf_clear(dbuffer[dst]);
/* Apply the filter, from dbuffer[src] to dbuffer[dst];
* if the filtering is canceled by the user mid-filter,
* we skip to the next filter without changing the source
* of the double buffering (so that the text goes through
* cleanly).
*/
if (filter(dbuffer[dst], dbuffer[src], filename) == 0) {
src = (src + 1) % 2;
}
if (git_buf_oom(dbuffer[dst]))
return GIT_ENOMEM;
}
/* Ensure that the output ends up in dbuffer[1] (i.e. the dest) */
if (dst != 1) {
git_buf_swap(dest, source);
}
return GIT_SUCCESS;
}
/*
* Copyright (C) 2009-2012 the libgit2 contributors
*
* This file is part of libgit2, distributed under the GNU GPL v2 with
* a Linking Exception. For full terms see the included COPYING file.
*/
#ifndef INCLUDE_filter_h__
#define INCLUDE_filter_h__
#include "common.h"
#include "buffer.h"
#include "git2/odb.h"
#include "git2/repository.h"
typedef int (*git_filter_cb)(git_buf *dest, const git_buf *source, const char *filename);
typedef enum {
GIT_FILTER_TO_WORKTREE,
GIT_FILTER_TO_ODB
} git_filter_mode;
typedef struct {
/* NUL, CR, LF and CRLF counts */
unsigned int nul, cr, lf, crlf;
/* These are just approximations! */
unsigned int printable, nonprintable;
} git_text_stats;
extern int git_filter__load_for_file(git_vector *filters, git_repository *repo, const char *full_path, int mode);
extern int git_filter__apply(git_buf *dest, git_buf *source, git_vector *filters, const char *filename);
/* Gather stats for a piece of text */
extern void git_text__stat(git_text_stats *stats, git_buf *text);
/* Heuristics on a set of text stats to check whether it's binary
* text or not */
extern int git_text__is_binary(git_text_stats *stats);
#endif
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment