Jump To …

general.c

libgit2 is a portable, pure C implementation of the Git core methods provided as a re-entrant linkable library with a solid API, allowing you to write native speed custom Git applications in any language which supports C bindings.

This file is an example of using that API in a real, compilable C file. As the API is updated, this file will be updated to demonstrate the new functionality.

If you're trying to write something in C using libgit2, you should also check out the generated API documentation. We try to link to the relevant sections of the API docs in each section in this file.

libgit2 (for the most part) only implements the core plumbing functions, not really the higher level porcelain stuff. For a primer on Git Internals that you will need to know to work with Git at this level, check out Chapter 9 of the Pro Git book.

Includes

 

      		
    

    

      
        

          
        

        
Including the git2.h header will include all the other libgit2 headers
that you need.  It should be the only thing you need to include in order
to compile properly and get all the libgit2 API.

      		
      
        
#include <git2.h>
#include <stdio.h>

      		
    

    

      
        

          
        

        
Almost all libgit2 functions return 0 on success or negative on error.
This is not production quality error checking, but should be sufficient
as an example.

      		
      
        
static void check_error(int error_code, const char *action)
{
  if (!error_code)
    return;

  const git_error *error = giterr_last();

  printf("Error %d %s - %s\n", error_code, action,
       (error && error->message) ? error->message : "???");

  exit(1);
}

int main (int argc, char** argv)
{

      		
    

    

      
        

          
        

        
Opening the Repository

      		
      
        

      		
    

    

      
        

          
        

        
There are a couple of methods for opening a repository, this being the
simplest.  There are also methods for specifying the index file
and work tree locations, here we assume they are in the normal places.



(Try running this program against tests-clar/resources/testrepo.git.)

      		
      
        
  int error;
  const char *repo_path = (argc > 1) ? argv[1] : "/opt/libgit2-test/.git";
  git_repository *repo;

  error = git_repository_open(&repo, repo_path);
  check_error(error, "opening repository");

      		
    

    

      
        

          
        

        
SHA-1 Value Conversions

      		
      
        

      		
    

    

      
        

          
        

        
For our first example, we will convert a 40 character hex value to the
20 byte raw SHA1 value.

      		
      
        
  printf("*Hex to Raw*\n");
  char hex[] = "4a202b346bb0fb0db7eff3cffeb3c70babbd2045";

      		
    

    

      
        

          
        

        
The git_oid is the structure that keeps the SHA value. We will use
this throughout the example for storing the value of the current SHA
key we're working with.

      		
      
        
  git_oid oid;
  git_oid_fromstr(&oid, hex);

      		
    

    

      
        

          
        

        
Once we've converted the string into the oid value, we can get the raw
value of the SHA by accessing oid.id

      		
      
        

      		
    

    

      
        

          
        

        
Next we will convert the 20 byte raw SHA1 value to a human readable 40
char hex value.

      		
      
        
  printf("\n*Raw to Hex*\n");
  char out[41];
  out[40] = '\0';

      		
    

    

      
        

          
        

        
If you have a oid, you can easily get the hex value of the SHA as well.

      		
      
        
  git_oid_fmt(out, &oid);
  printf("SHA hex string: %s\n", out);

      		
    

    

      
        

          
        

        
Working with the Object Database

      		
      
        

      		
    

    

      
        

          
        

        
libgit2 provides direct access to the object database.  The
object database is where the actual objects are stored in Git. For
working with raw objects, we'll need to get this structure from the
repository.

      		
      
        
  git_odb *odb;
  git_repository_odb(&odb, repo);

      		
    

    

      
        

          
        

        
Raw Object Reading

      		
      
        
  printf("\n*Raw Object Read*\n");
  git_odb_object *obj;
  git_otype otype;
  const unsigned char *data;
  const char *str_type;

      		
    

    

      
        

          
        

        
We can read raw objects directly from the object database if we have
the oid (SHA) of the object.  This allows us to access objects without
knowing thier type and inspect the raw bytes unparsed.

      		
      
        
  error = git_odb_read(&obj, odb, &oid);
  check_error(error, "finding object in repository");

      		
    

    

      
        

          
        

        
A raw object only has three properties - the type (commit, blob, tree
or tag), the size of the raw data and the raw, unparsed data itself.
For a commit or tag, that raw data is human readable plain ASCII
text. For a blob it is just file contents, so it could be text or
binary data. For a tree it is a special binary format, so it's unlikely
to be hugely helpful as a raw object.

      		
      
        
  data = (const unsigned char *)git_odb_object_data(obj);
  otype = git_odb_object_type(obj);

      		
    

    

      
        

          
        

        
We provide methods to convert from the object type which is an enum, to
a string representation of that value (and vice-versa).

      		
      
        
  str_type = git_object_type2string(otype);
  printf("object length and type: %d, %s\n",
      (int)git_odb_object_size(obj),
      str_type);

      		
    

    

      
        

          
        

        
For proper memory management, close the object when you are done with
it or it will leak memory.

      		
      
        
  git_odb_object_free(obj);

      		
    

    

      
        

          
        

        
Raw Object Writing

      		
      
        
  printf("\n*Raw Object Write*\n");

      		
    

    

      
        

          
        

        
You can also write raw object data to Git. This is pretty cool because
it gives you direct access to the key/value properties of Git.  Here
we'll write a new blob object that just contains a simple string.
Notice that we have to specify the object type as the git_otype enum.

      		
      
        
  git_odb_write(&oid, odb, "test data", sizeof("test data") - 1, GIT_OBJ_BLOB);

      		
    

    

      
        

          
        

        
Now that we've written the object, we can check out what SHA1 was
generated when the object was written to our database.

      		
      
        
  git_oid_fmt(out, &oid);
  printf("Written Object: %s\n", out);

      		
    

    

      
        

          
        

        
Object Parsing

      		
      
        

      		
    

    

      
        

          
        

        
libgit2 has methods to parse every object type in Git so you don't have
to work directly with the raw data. This is much faster and simpler
than trying to deal with the raw data yourself.

      		
      
        

      		
    

    

      
        

          
        

        
Commit Parsing

      		
      
        

      		
    

    

      
        

          
        

        
Parsing commit objects is simple and gives you access to all the
data in the commit - the author (name, email, datetime), committer
(same), tree, message, encoding and parent(s).

      		
      
        
  printf("\n*Commit Parsing*\n");

  git_commit *commit;
  git_oid_fromstr(&oid, "8496071c1b46c854b31185ea97743be6a8774479");

  error = git_commit_lookup(&commit, repo, &oid);
  check_error(error, "looking up commit");

  const git_signature *author, *cmtter;
  const char *message;
  time_t ctime;
  unsigned int parents, p;

      		
    

    

      
        

          
        

        
Each of the properties of the commit object are accessible via methods,
including commonly needed variations, such as git_commit_time which
returns the author time and git_commit_message which gives you the
commit message (as a NUL-terminated string).

      		
      
        
  message  = git_commit_message(commit);
  author   = git_commit_author(commit);
  cmtter   = git_commit_committer(commit);
  ctime    = git_commit_time(commit);

      		
    

    

      
        

          
        

        
The author and committer methods return [gitsignature] structures,
which give you name, email and when, which is a `gittime` structure,
giving you a timestamp and timezone offset.

      		
      
        
  printf("Author: %s (%s)\n", author->name, author->email);

      		
    

    

      
        

          
        

        
Commits can have zero or more parents. The first (root) commit will
have no parents, most commits will have one (i.e. the commit it was
based on) and merge commits will have two or more.  Commits can
technically have any number, though it's rare to have more than two.

      		
      
        
  parents  = git_commit_parentcount(commit);
  for (p = 0;p < parents;p++) {
    git_commit *parent;
    git_commit_parent(&parent, commit, p);
    git_oid_fmt(out, git_commit_id(parent));
    printf("Parent: %s\n", out);
    git_commit_free(parent);
  }

      		
    

    

      
        

          
        

        
Don't forget to close the object to prevent memory leaks. You will have
to do this for all the objects you open and parse.

      		
      
        
  git_commit_free(commit);

      		
    

    

      
        

          
        

        
Writing Commits

      		
      
        

      		
    

    

      
        

          
        

        
libgit2 provides a couple of methods to create commit objects easily as
well. There are four different create signatures, we'll just show one
of them here.  You can read about the other ones in the commit API
docs.

      		
      
        
  printf("\n*Commit Writing*\n");
  git_oid tree_id, parent_id, commit_id;
  git_tree *tree;
  git_commit *parent;

      		
    

    

      
        

          
        

        
Creating signatures for an authoring identity and time is simple.  You
will need to do this to specify who created a commit and when.  Default
values for the name and email should be found in the user.name and
user.email configuration options.  See the config section of this
example file to see how to access config values.

      		
      
        
  git_signature_new((git_signature **)&author,
      "Scott Chacon", "schacon@gmail.com", 123456789, 60);
  git_signature_new((git_signature **)&cmtter,
      "Scott A Chacon", "scott@github.com", 987654321, 90);

      		
    

    

      
        

          
        

        
Commit objects need a tree to point to and optionally one or more
parents.  Here we're creating oid objects to create the commit with,
but you can also use

      		
      
        
  git_oid_fromstr(&tree_id, "f60079018b664e4e79329a7ef9559c8d9e0378d1");
  git_tree_lookup(&tree, repo, &tree_id);
  git_oid_fromstr(&parent_id, "5b5b025afb0b4c913b4c338a42934a3863bf3644");
  git_commit_lookup(&parent, repo, &parent_id);

      		
    

    

      
        

          
        

        
Here we actually create the commit object with a single call with all
the values we need to create the commit.  The SHA key is written to the
commit_id variable here.

      		
      
        
  git_commit_create_v(
    &commit_id, /* out id */
    repo,
    NULL, /* do not update the HEAD */
    author,
    cmtter,
    NULL, /* use default message encoding */
    "example commit",
    tree,
    1, parent);

      		
    

    

      
        

          
        

        
Now we can take a look at the commit SHA we've generated.

      		
      
        
  git_oid_fmt(out, &commit_id);
  printf("New Commit: %s\n", out);

      		
    

    

      
        

          
        

        
Tag Parsing

      		
      
        

      		
    

    

      
        

          
        

        
You can parse and create tags with the tag management API, which
functions very similarly to the commit lookup, parsing and creation
methods, since the objects themselves are very similar.

      		
      
        
  printf("\n*Tag Parsing*\n");
  git_tag *tag;
  const char *tmessage, *tname;
  git_otype ttype;

      		
    

    

      
        

          
        

        
We create an oid for the tag object if we know the SHA and look it up
the same way that we would a commit (or any other object).

      		
      
        
  git_oid_fromstr(&oid, "b25fa35b38051e4ae45d4222e795f9df2e43f1d1");

  error = git_tag_lookup(&tag, repo, &oid);
  check_error(error, "looking up tag");

      		
    

    

      
        

          
        

        
Now that we have the tag object, we can extract the information it
generally contains: the target (usually a commit object), the type of
the target object (usually 'commit'), the name ('v1.0'), the tagger (a
git_signature - name, email, timestamp), and the tag message.

      		
      
        
  git_tag_target((git_object **)&commit, tag);
  tname = git_tag_name(tag);		// "test"
  ttype = git_tag_target_type(tag);	// GIT_OBJ_COMMIT (otype enum)
  tmessage = git_tag_message(tag);	// "tag message\n"
  printf("Tag Message: %s\n", tmessage);

  git_commit_free(commit);

      		
    

    

      
        

          
        

        
Tree Parsing

      		
      
        

      		
    

    

      
        

          
        

        
Tree parsing is a bit different than the other objects, in that
we have a subtype which is the tree entry.  This is not an actual
object type in Git, but a useful structure for parsing and traversing
tree entries.

      		
      
        
  printf("\n*Tree Parsing*\n");

  const git_tree_entry *entry;
  git_object *objt;

      		
    

    

      
        

          
        

        
Create the oid and lookup the tree object just like the other objects.

      		
      
        
  git_oid_fromstr(&oid, "2a741c18ac5ff082a7caaec6e74db3075a1906b5");
  git_tree_lookup(&tree, repo, &oid);

      		
    

    

      
        

          
        

        
Getting the count of entries in the tree so you can iterate over them
if you want to.

      		
      
        
  size_t cnt = git_tree_entrycount(tree); // 3
  printf("tree entries: %d\n", (int)cnt);

  entry = git_tree_entry_byindex(tree, 0);
  printf("Entry name: %s\n", git_tree_entry_name(entry)); // "hello.c"

      		
    

    

      
        

          
        

        
You can also access tree entries by name if you know the name of the
entry you're looking for.

      		
      
        
  entry = git_tree_entry_byname(tree, "README");
  git_tree_entry_name(entry); // "hello.c"

      		
    

    

      
        

          
        

        
Once you have the entry object, you can access the content or subtree
(or commit, in the case of submodules) that it points to.  You can also
get the mode if you want.

      		
      
        
  git_tree_entry_to_object(&objt, repo, entry); // blob

      		
    

    

      
        

          
        

        
Remember to close the looked-up object once you are done using it

      		
      
        
  git_object_free(objt);

      		
    

    

      
        

          
        

        
Blob Parsing

      		
      
        

      		
    

    

      
        

          
        

        
The last object type is the simplest and requires the least parsing
help. Blobs are just file contents and can contain anything, there is
no structure to it. The main advantage to using the simple blob
api is that when you're creating blobs you don't have to calculate
the size of the content.  There is also a helper for reading a file
from disk and writing it to the db and getting the oid back so you
don't have to do all those steps yourself.

      		
      
        
  printf("\n*Blob Parsing*\n");
  git_blob *blob;

  git_oid_fromstr(&oid, "1385f264afb75a56a5bec74243be9b367ba4ca08");
  git_blob_lookup(&blob, repo, &oid);

      		
    

    

      
        

          
        

        
You can access a buffer with the raw contents of the blob directly.
Note that this buffer may not be contain ASCII data for certain blobs
(e.g. binary files): do not consider the buffer a NULL-terminated
string, and use the git_blob_rawsize attribute to find out its exact
size in bytes

      		
      
        
  printf("Blob Size: %ld\n", (long)git_blob_rawsize(blob)); // 8
  git_blob_rawcontent(blob); // "content"

      		
    

    

      
        

          
        

        
Revwalking

      		
      
        

      		
    

    

      
        

          
        

        
The libgit2 revision walking api provides methods to traverse the
directed graph created by the parent pointers of the commit objects.
Since all commits point back to the commit that came directly before
them, you can walk this parentage as a graph and find all the commits
that were ancestors of (reachable from) a given starting point.  This
can allow you to create git log type functionality.

      		
      
        
  printf("\n*Revwalking*\n");
  git_revwalk *walk;
  git_commit *wcommit;

  git_oid_fromstr(&oid, "5b5b025afb0b4c913b4c338a42934a3863bf3644");

      		
    

    

      
        

          
        

        
To use the revwalker, create a new walker, tell it how you want to sort
the output and then push one or more starting points onto the walker.
If you want to emulate the output of git log you would push the SHA
of the commit that HEAD points to into the walker and then start
traversing them.  You can also 'hide' commits that you want to stop at
or not see any of their ancestors.  So if you want to emulate git log
branch1..branch2, you would push the oid of branch2 and hide the oid
of branch1.

      		
      
        
  git_revwalk_new(&walk, repo);
  git_revwalk_sorting(walk, GIT_SORT_TOPOLOGICAL | GIT_SORT_REVERSE);
  git_revwalk_push(walk, &oid);

  const git_signature *cauth;
  const char *cmsg;

      		
    

    

      
        

          
        

        
Now that we have the starting point pushed onto the walker, we start
asking for ancestors. It will return them in the sorting order we asked
for as commit oids.  We can then lookup and parse the commited pointed
at by the returned OID; note that this operation is specially fast
since the raw contents of the commit object will be cached in memory

      		
      
        
  while ((git_revwalk_next(&oid, walk)) == 0) {
    error = git_commit_lookup(&wcommit, repo, &oid);
  check_error(error, "looking up commit during revwalk");

    cmsg  = git_commit_message(wcommit);
    cauth = git_commit_author(wcommit);
    printf("%s (%s)\n", cmsg, cauth->email);

    git_commit_free(wcommit);
  }

      		
    

    

      
        

          
        

        
Like the other objects, be sure to free the revwalker when you're done
to prevent memory leaks.  Also, make sure that the repository being
walked it not deallocated while the walk is in progress, or it will
result in undefined behavior

      		
      
        
  git_revwalk_free(walk);

      		
    

    

      
        

          
        

        
Index File Manipulation

      		
      
        

      		
    

    

      
        

          
        

        
The index file API allows you to read, traverse, update and write
the Git index file (sometimes thought of as the staging area).

      		
      
        
  printf("\n*Index Walking*\n");

  git_index *index;
  unsigned int i, ecount;

      		
    

    

      
        

          
        

        
You can either open the index from the standard location in an open
repository, as we're doing here, or you can open and manipulate any
index file with git_index_open_bare(). The index for the repository
will be located and loaded from disk.

      		
      
        
  git_repository_index(&index, repo);

      		
    

    

      
        

          
        

        
For each entry in the index, you can get a bunch of information
including the SHA (oid), path and mode which map to the tree objects
that are written out.  It also has filesystem properties to help
determine what to inspect for changes (ctime, mtime, dev, ino, uid,
gid, filesize and flags) All these properties are exported publicly in
the `gitindex_entry` struct

      		
      
        
  ecount = git_index_entrycount(index);
  for (i = 0; i < ecount; ++i) {
    const git_index_entry *e = git_index_get_byindex(index, i);

    printf("path: %s\n", e->path);
    printf("mtime: %d\n", (int)e->mtime.seconds);
    printf("fs: %d\n", (int)e->file_size);
  }

  git_index_free(index);

      		
    

    

      
        

          
        

        
References

      		
      
        

      		
    

    

      
        

          
        

        
The reference API allows you to list, resolve, create and update
references such as branches, tags and remote references (everything in
the .git/refs directory).

      		
      
        
  printf("\n*Reference Listing*\n");

      		
    

    

      
        

          
        

        
Here we will implement something like git for-each-ref simply listing
out all available references and the object SHA they resolve to.

      		
      
        
  git_strarray ref_list;
  git_reference_list(&ref_list, repo, GIT_REF_LISTALL);

  const char *refname;
  git_reference *ref;

      		
    

    

      
        

          
        

        
Now that we have the list of reference names, we can lookup each ref
one at a time and resolve them to the SHA, then print both values out.

      		
      
        
  for (i = 0; i < ref_list.count; ++i) {
    refname = ref_list.strings[i];
    git_reference_lookup(&ref, repo, refname);

    switch (git_reference_type(ref)) {
    case GIT_REF_OID:
      git_oid_fmt(out, git_reference_target(ref));
      printf("%s [%s]\n", refname, out);
      break;

    case GIT_REF_SYMBOLIC:
      printf("%s => %s\n", refname, git_reference_symbolic_target(ref));
      break;
    default:
      fprintf(stderr, "Unexpected reference type\n");
      exit(1);
    }
  }

  git_strarray_free(&ref_list);

      		
    

    

      
        

          
        

        
Config Files

      		
      
        

      		
    

    

      
        

          
        

        
The config API allows you to list and updatee config values
in any of the accessible config file locations (system, global, local).

      		
      
        
  printf("\n*Config Listing*\n");

  const char *email;
  int32_t j;

  git_config *cfg;

      		
    

    

      
        

          
        

        
Open a config object so we can read global values from it.

      		
      
        
  char config_path[256];
  sprintf(config_path, "%s/config", repo_path);
  check_error(git_config_open_ondisk(&cfg, config_path), "opening config");

  git_config_get_int32(&j, cfg, "help.autocorrect");
  printf("Autocorrect: %d\n", j);

  git_config_get_string(&email, cfg, "user.email");
  printf("Email: %s\n", email);

      		
    

    

      
        

          
        

        
Finally, when you're done with the repository, you can free it as well.


      		
      
        
  git_repository_free(repo);

  return 0;
}