git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 6/6] diff.c: detect blocks despite whitespace changes

From
Junio C Hamano <gitster@pobox.com>
Date
Jun 28, 2017, 05:06 UTC
Message-ID
<xmqq37akya3j.fsf@gitster.mtv.corp.google.com>
In-Reply-To
<xmqqshikye0o.fsf@gitster.mtv.corp.google.com>
Junio C Hamano <gitster@pobox.com> writes:
Show 5 quoted lines
> Looking at the implementation of get_ws_cleaned_string() that is the
> workhorse of emitted_symbol_cmp_no_ws(), it seems to be doing wrong
> things for various "ignore whitespace" options (i.e. there is only
> one implementation, while "git diff" family takes things like
> "ignore space change", "ignore all whitespace", etc.), though.

This probably deserves a bit more illustration of how I envision the code should evolve.

In the longer term, I would prefer to see emitted_symbol_cmp_no_ws() to go and instead emitted_symbol_cmp() to take the diff options so that it can change the behaviour of the comparison function based on the -w/-b/--ignore-space-at-eol/etc. settings. And compare two strings in place.

For that, you may need a helper function that takes a pointer to a character pointer, picks the next byte that matters while advancing the pointer, and returns that byte. The emitted_symbol_cmp(a, b) which is not used for real comparison (i.e. ordering to see if a sorts earlier than b) but for equivalence (i.e. considering various whitespace-ignoring settings, does a and b matfch?) may become something like:

        int
        emitted_symbol_eqv(struct emitted_diff_symbol *a,
                           struct emitted_diff_symbol *b,
                           const void *keydata) {
                struct diff_options *diffopt = keydata;
                const char *ap = a->line;
                const char *bp = b->line;
                while (1) {
                        int ca, cb;
                        ca = next_byte(&ap, diffopt);
                        cb = next_byte(&bp, diffopt);
                        if (ca != cb)
                                return 1; /* differs */
                        if (!ca)
                                return 0;
                };
        }                           
where next_byte() may look like:
        static int
        next_byte(const char **cp, struct diff_options *diffopt)
        {                  
                int retval;
        again:
                retval = **cp;
                if (!retval)
                        return retval; /* end of string */
                if (!isspace(retval)) {
                        (*cp)++; /* advance */
                        return retval;
                }
                switch (ignore_space_type(diffopt)) {
                case NOT_IGNORING:
                        (*cp)++; /* advance */
                        return retval;
                case IGNORE_SPACE_CHANGE:
                        while (**cp && isspace(**cp))
                                (*cp)++; /* squash consecutive spaces */
                        return ' '; /* normalize spaces with a SP */
                case IGNORE_ALL_SPACE:
                        (*cp)++; /* advance */
                        goto again;
                ... other cases here ...
                }
        }
Previous: Junio C HamanoNext: Stefan Beller
Message 14 of 18 in “Fixing up sb/diff-color-moved”
  1. 0/6 Fixing up sb/diff-color-movedStefan Beller, Jun 28, 2017
  2. 1/6 diff.c: factor out shrinking of potential moved line blocksStefan Beller, Jun 28, 2017
  3. 2/6 diff.c: change the default for move coloring to zebraStefan Beller, Jun 28, 2017
  4. Junio C HamanoJun 28, 2017
  5. Stefan BellerJun 28, 2017
  6. 3/6 diff.c: better reporting on color.moved bogus configurationStefan Beller, Jun 28, 2017
  7. 4/6 Documentation/diff: reword color movedStefan Beller, Jun 28, 2017
  8. Junio C HamanoJun 28, 2017
  9. 5/6 diff.c: omit uninteresting moved linesStefan Beller, Jun 28, 2017
  10. Junio C HamanoJun 28, 2017
  11. Stefan BellerJun 28, 2017
  12. 6/6 diff.c: detect blocks despite whitespace changesStefan Beller, Jun 28, 2017
  13. Junio C HamanoJun 28, 2017
  14. Junio C HamanoJun 28, 2017
  15. Stefan BellerJun 28, 2017
  16. Junio C HamanoJun 28, 2017
  17. Stefan BellerJun 29, 2017
  18. Junio C HamanoJun 29, 2017

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.