Am 14.11.20 um 18:35 schrieb Martin Ågren: > Hi, > > On Sat, 14 Nov 2020 at 16:55, Grégoire PARIS <postmaster@xxxxxxxxxxx> wrote: > >>> In the end I think your best option right now is to give explicit line >>> numbers for <end> and <start>. >> That is indeed what I currently do, I plugged that to vim's visual >> selection with >> >> vnoremap <leader>l :<c-u>exe '!git log -L' >> line("'<").','.line("'>").':'.expand('%')<CR> >> >> and it works great! > > Great! > >> I also suppose the issue is the same for any other language that has >> documentation above function declarations. > > Yeah, you'll see the same thing for C files, e.g., using this in the > repo of Git itself: > > git log -L :strbuf_swap:strbuf.h > > It will follow the lines from the function definition all the way down > to the next function, but just as you saw in your example, it will not > match the comment immediately *before* the function. That is, these > lines will be followed: > > https://github.com/git/git/blob/v2.29.2/strbuf.h#L125-L138 The --function-context options of git diff and git grep try to show comments by including non-empty lines before function lines. This heuristic might work for -L :funcname:file as well (patch below), but breaks seven tests in each of t8001-annotate.sh, t8002-blame.sh and t8012-blame-colors.sh. René --- line-range.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/line-range.c b/line-range.c index 9b50583dc0..5d55fa255f 100644 --- a/line-range.c +++ b/line-range.c @@ -163,6 +163,13 @@ static const char *find_funcname_matching_regexp(xdemitconf_t *xecfg, const char } } +static int is_empty_line(const char *bol, const char *eol) +{ + while (bol < eol && isspace(*bol)) + bol++; + return bol == eol; +} + static const char *parse_range_funcname( const char *arg, nth_line_fn_t nth_line_cb, void *cb_data, long lines, long anchor, long *begin, long *end, @@ -233,6 +240,18 @@ static const char *parse_range_funcname( (*end)++; } + /* + * Include non-empty, non-funcname lines before the found + * funcname line, as they probably contain related comments. + */ + while (*begin > 0) { + const char *bol = nth_line_cb(cb_data, *begin - 1); + const char *eol = nth_line_cb(cb_data, *begin); + if (is_empty_line(bol, eol) || match_funcname(xecfg, bol, eol)) + break; + (*begin)--; + } + regfree(®exp); free(xecfg); free(pattern); -- 2.29.2