Features/diff generator

[cholin]

generator implementation for diff (#183)

[arrbee]

@cholin How did that patch part of the diff API work for you? Was is documented clearly enough? It's still relatively new, so I'm happy for any feedback.

For example, I've considered adding an API that let's you run linearly over the lines in the patch instead of requiring nested loops, a la:

typedef struct {
    git_diff_range range;
    const char *header;
    size_t header_len;
    size_t hunk_of_patch;
    size_t lines_in_hunk;
} git_diff_hunk;

typedef struct {
    char line_origin;
    const char *content;
    size_t content_len;
    size_t line_of_hunk;
    int old_lineno;
    int new_lineno;
} git_diff_line;

GIT_EXTERN(size_t) git_diff_patch_num_lines(
    const git_diff_patch *patch);

GIT_EXTERN(int) git_diff_patch_get_line(
    const git_diff_hunk **hunk_out,
    const git_diff_line **line_out,
    const git_diff_patch *patch,
    size_t line);

You could look at the line_of_hunk field in the line object to know when you were starting a new hunk (and the hunk_out pointer had changed).

Anyhow, I don't know that this is necessarily a good idea, but since you've been using the API recently and since it hasn't seen much use yet, I thought I'd try to solicit some feedback. Thanks!

[cholin]

@arrbee: Documentation was good! I found all the necessary informations through the api documentation. I will give you a more detailed feedback in a few days. Currently I'm short on time. ;)

[cholin]

@arrbee: Ok I missed the bus, so here are my thoughts about diff in libgit2

The documentation in the diff.h is really good but unfortunately all constants are missing at libgit2.github.com/libgit2 (like GIT_DELTA_* or GIT_DIFF__). For git_diff_find_similiar I had problems with too small files (error msg: *File too small for similarity signature calculation_). The message is clear but I had to dig deep into the code to get the minimum size to do a similarity signature calculation. If I'm right it's 127 (defined in hashsig.c - HASHSIG_HEAP_SIZE). As well I thought that the check aborts with this error msg ( but that was just a feeling ;) ).

I think your consideration about extending the diff api to run linearly over the lines is a good idea. With that it's really easy to get all the informations in a flat iteration style. But I don't know if it's perhaps confusing for the user
with all these possiblities to get the diff informations.

Overall I think you did a great job with this api! 👍