Function processLog

programs/git-import/git-import.cpp:1142–1196

Obtain the list of commits and process them. */

/** Obtain the list of commits and process them.
  */
static void processLog(const Options & options)
{
    ResultWriter result;

    std::string command = "git log --reverse --no-merges --pretty=%H";
    fmt::print("{}\n", command);
    auto git_log = ShellCommand::execute(command);

    /// Collect hashes in memory. This is inefficient but allows to display beautiful progress.
    /// The number of commits is in order of single millions for the largest repositories,
    /// so don't care about potential waste of ~100 MB of memory.

    std::vector<std::string> hashes;

    auto & in = git_log->out;
    while (!in.eof())
    {
        std::string hash;
        readString(hash, in);
        assertChar('\n', in);

        if (!options.skip_commits.contains(hash))
            hashes.emplace_back(std::move(hash));
    }

    size_t num_commits = hashes.size();
    fmt::print("Total {} commits to process.\n", num_commits);

    /// Will run multiple processes in parallel
    size_t num_threads = options.threads;
    if (num_threads == 0)
        throw Exception(ErrorCodes::INCORRECT_DATA, "num-threads cannot be zero");

    std::vector<std::unique_ptr<ShellCommand>> show_commands(num_threads);
    for (size_t i = 0; i < num_commits && i < num_threads; ++i)
        show_commands[i] = gitShow(hashes[i]);

    Snapshot snapshot;
    DiffHashes diff_hashes;

    for (size_t i = 0; i < num_commits; ++i)
    {
        ReadBuffer & show_in = show_commands[i % num_threads]->out;
        processCommit(show_in, options, i, num_commits, hashes[i], snapshot, diff_hashes, result);
        show_in.ignoreAll();

        if (!options.stop_after_commit.empty() && hashes[i] == options.stop_after_commit)
            break;

        if (i + num_threads < num_commits)
            show_commands[i % num_threads] = gitShow(hashes[i + num_threads]);
    }

    result.finalize();
}


}