Week 3: A Dubious Approach!

Hi everyone,

Its almost a month since I have been selected as GSoC Student by Git. I explored and tried various stuff within git source code to think of a particular approach for my project.

Last week, I studied the flow of data within git log and git for-each-ref. I think, to start (re)using ref-filter’s functions, pretty.c must need to understand ref-filter’s atoms… And to understand atoms, pretty.c must need to support ref-filter’s structures.

Here are the few structures used by pretty.c

cmt_fmt_map

• Cannot stop using this, but good candidate to make transition table

pretty_print_context

• All pp_* functions depends on this. But I guess some fields can be ditched in favour of using ref-filter’s [might need to modify pp_* functions too]

userformat_want

• No equivalance present in ref-filter (Not even play much role in formatting logic)

Whats next?

I guess, I’m going to focus more on structures.

I see many part of ‘git log’ are using struct rev_info *revs and funtions are filling values in revs.

Its time to introduce struct ref_array to ‘pretty.c’. So, we can start filling values in array.revs instead of revs.

struct ref_array {
	int nr, alloc;
	struct ref_array_item **items;
	struct rev_info *revs;
} array;

This might be a good start point.