| Message ID | pull.1236.git.1653310466062.gitgitgadget@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | check-ignore: --non-matching without --verbose | expand |
"Carl Smedstad via GitGitGadget" <gitgitgadget@gmail.com> writes: > The motivation for this is to use the underutilized invocation of > --non-matching without --verbose to create a shorthand for what could be > achieved by: Not commenting on the patch itself, but ... > > find . | > git check-ignore --verbose --non-matching | > grep $'^::\t' | > sed -e 's/.*\t//' ... feeding grep output to sed caught my eyes. Isn't that equivalent to git ... | sed -n '/^::\t/s/.*\t//p' ?
On 5/23/2022 8:54 AM, Carl Smedstad via GitGitGadget wrote: > From: Carl Smedstad <carl.smedstad@protonmail.com> > -n, --non-matching:: > - Show given paths which don't match any pattern. This only > - makes sense when `--verbose` is enabled, otherwise it would > - not be possible to distinguish between paths which match a > - pattern and those which don't. > + Only show given paths which don't match any pattern. If `--verbose` is > + enabled, show both paths that match a pattern and those which don't. This may be better to indicate the behavior as predicated on the existence of --verbose: If `--verbose` is enabled, then all paths are listed along with an indicator (`::`) that no matching pattern was found. Without `--verbose`, list only the paths that do not match any pattern. > - if (!quiet && (pattern || show_non_matching)) > - output_pattern(pathspec.items[i].original, pattern); > + if (!quiet) { > + if (verbose) { > + if (show_non_matching || pattern) > + output_pattern(pathspec.items[i].original, pattern); > + } else { > + if (show_non_matching && !pattern) > + output_pattern(pathspec.items[i].original, pattern); > + if (!show_non_matching && pattern) > + output_pattern(pathspec.items[i].original, pattern); These three blocks all call the same code line. So really you want to avoid a single case: if (!quiet && ((verbose && (show_non_matching || pattern)) || (!verbose && !!show_non_matching != !!pattern))) This is the most direct way to write what you had above. However, we could do this more simply: /* If --non-matching, then show if verbose or the pattern is missing. */ if (!quiet && show_non_matching && (verbose || !pattern)) output_pattern(...); /* If not --non-matching, then show if the pattern exists. */ if (!quiet && !show_non_matching && pattern) output_pattern(...); Hopefully that's a bit easier to parse. I believe it is equivalent. Thanks, -Stolee
> Not commenting on the patch itself, but ... > > > find . | > > git check-ignore --verbose --non-matching | > > grep $'^::\t' | > > sed -e 's/.*\t//' > > ... feeding grep output to sed caught my eyes. Isn't that equivalent to > > git ... | > sed -n '/^::\t/s/.*\t//p' That is indeed equivalent. TIL about regular-expression addresses in sed, thanks!
> This may be better to indicate the behavior as predicated on the > existence of --verbose: > > If `--verbose` is enabled, then all paths are listed along > with an indicator (`::`) that no matching pattern was found. > Without `--verbose`, list only the paths that do not match > any pattern. Good point. I've adopted you changes but made some slight changes to the first line: If `--verbose` is enabled, list both matching and non-matching paths (non-matching paths along with the indicator `::`). Without `--verbose`, list only the paths that do not match any pattern. > These three blocks all call the same code line. So really you want > to avoid a single case: > > if (!quiet && > ((verbose && (show_non_matching || pattern)) || > (!verbose && !!show_non_matching != !!pattern))) > > This is the most direct way to write what you had above. However, > we could do this more simply: > > /* If --non-matching, then show if verbose or the pattern is missing. */ > if (!quiet && show_non_matching && (verbose || !pattern)) > output_pattern(...); > > /* If not --non-matching, then show if the pattern exists. */ > if (!quiet && !show_non_matching && pattern) > output_pattern(...); > > Hopefully that's a bit easier to parse. I believe it is > equivalent. That is indeed equivalent and a lot easier to read. Implemented the changes in full. Thanks!
"Carl Smedstad via GitGitGadget" <gitgitgadget@gmail.com> writes: > -n, --non-matching:: > - Show given paths which don't match any pattern. This only > - makes sense when `--verbose` is enabled, otherwise it would > - not be possible to distinguish between paths which match a > - pattern and those which don't. OK. In the current code, verbose mode output can show both the pattern that decided the fate for the path and the path itself, so we can afford to include paths that match and that do not match. There may not be any pattern that matched the path, or the last pattern that matched the path was negative, or the last pattern that matched was positive. Users can tell them apart (the first two cases are the paths that do not match, the last case is the paths that do match). But non-verbose mode only show paths, so showing both matched and non-matching paths to the output would not make sense. But we document that it is a silly thing to ask, and explicitly reject the combination of non-verbose option and --non-matching option. So the option is misnamed; it is more like 'include-non-matching', not 'non-matching-only'. > + Only show given paths which don't match any pattern. If `--verbose` is > + enabled, show both paths that match a pattern and those which don't. But this patch proposes the option to 'non-matching-only' only in non-verbose mode. Since "check-ignore --non-matching <pathspec>" without "--verbose" immediately errors out, making it do something different does not have too much backward-compatibility worries. It may be confusing to the users that the set of paths included in the output, when the same "-n" option is given, were different between the non-verbose and the verbose mode. That is one thing I feel uncomfortable about this change. I wonder if we are better off to invent a new "--non-matching-only" option, have it support both the non-verbose and the verbose mode, and leave the current "-n" alone, with an option to perhaps giving "--include-non-matching" alias to the latter to clarify? > diff --git a/builtin/check-ignore.c b/builtin/check-ignore.c > index 21912569650..10e8bba1057 100644 > --- a/builtin/check-ignore.c > +++ b/builtin/check-ignore.c > @@ -113,8 +113,17 @@ static int check_ignore(struct dir_struct *dir, > pattern->flags & PATTERN_FLAG_NEGATIVE) > pattern = NULL; > } > - if (!quiet && (pattern || show_non_matching)) > - output_pattern(pathspec.items[i].original, pattern); > + if (!quiet) { > + if (verbose) { > + if (show_non_matching || pattern) > + output_pattern(pathspec.items[i].original, pattern); > + } else { > + if (show_non_matching && !pattern) > + output_pattern(pathspec.items[i].original, pattern); > + if (!show_non_matching && pattern) > + output_pattern(pathspec.items[i].original, pattern); > + } > + } > if (pattern) > num_ignored++; > } > @@ -175,8 +184,6 @@ int cmd_check_ignore(int argc, const char **argv, const char *prefix) > if (verbose) > die(_("cannot have both --quiet and --verbose")); > } > - if (show_non_matching && !verbose) > - die(_("--non-matching is only valid with --verbose")); > > /* read_cache() is only necessary so we can watch out for submodules. */ > if (!no_index && read_cache() < 0) > diff --git a/t/t0008-ignores.sh b/t/t0008-ignores.sh > index 5575dade8ee..2ea178797bc 100755 > --- a/t/t0008-ignores.sh > +++ b/t/t0008-ignores.sh > @@ -90,7 +90,8 @@ test_check_ignore () { > # 1. with -q / --quiet > # 2. with default verbosity > # 3. with -v / --verbose > -# 4. with -v / --verbose, *and* -n / --non-matching > +# 4. with -n / --non-matching > +# 5. with -v / --verbose, *and* -n / --non-matching > # > # expecting success each time. Takes advantage of the fact that > # check-ignore --verbose output is the same as normal output except > @@ -102,7 +103,7 @@ test_check_ignore () { > # Arguments: > # - (optional) prereqs for this test, e.g. 'SYMLINKS' > # - test name > -# - output to expect from the fourth verbosity mode (the output > +# - output to expect from the fifth verbosity mode (the output > # from the other verbosity modes is automatically inferred > # from this value) > # - code to run (should invoke test_check_ignore) > @@ -124,6 +125,7 @@ test_expect_success_multiple () { > > expect_verbose=$( echo "$expect_all" | grep -v '^:: ' ) > expect=$( echo "$expect_verbose" | sed -e 's/.* //' ) > + expect_non_matching=$( echo "$expect_all" | grep '^:: ' | sed -e 's/.* //') > > test_expect_success $prereq "$testname${no_index_opt:+ with $no_index_opt}" ' > expect "$expect" && > @@ -144,6 +146,17 @@ test_expect_success_multiple () { > quiet_opt= > fi > > + for non_matching_opt in '-n' '--non-matching' > + do > + test_code=" > + expect '$expect_non_matching' && > + $code > + " > + opts="${no_index_opt:+$no_index_opt }$non_matching_opt" > + test_expect_success $prereq "$testname${opts:+ with $opts}" "$test_code" > + done > + non_matching_opt= > + > for verbose_opt in '-v' '--verbose' > do > for non_matching_opt in '' '-n' '--non-matching' > > base-commit: f9b95943b68b6b8ca5a6072f50a08411c6449b55
Carl Smedstad <carl.smedstad@protonmail.com> writes: >> Not commenting on the patch itself, but ... >> >> > find . | >> > git check-ignore --verbose --non-matching | >> > grep $'^::\t' | >> > sed -e 's/.*\t//' >> >> ... feeding grep output to sed caught my eyes. Isn't that equivalent to >> >> git ... | >> sed -n '/^::\t/s/.*\t//p' > > That is indeed equivalent. TIL about regular-expression addresses in > sed, thanks! I however suspect that these "\t"'s should be spelled out as an actual TAB in the final production version for portability. Thanks.
On Tue, May 24 2022, Junio C Hamano wrote: > Carl Smedstad <carl.smedstad@protonmail.com> writes: > >>> Not commenting on the patch itself, but ... >>> >>> > find . | >>> > git check-ignore --verbose --non-matching | >>> > grep $'^::\t' | >>> > sed -e 's/.*\t//' >>> >>> ... feeding grep output to sed caught my eyes. Isn't that equivalent to >>> >>> git ... | >>> sed -n '/^::\t/s/.*\t//p' >> >> That is indeed equivalent. TIL about regular-expression addresses in >> sed, thanks! > > I however suspect that these "\t"'s should be spelled out as an > actual TAB in the final production version for portability. Make that a "for sure", it's a commonly unsupported feature of some sed's, it's actually more common that not that it isn't supported. See $HT in t4213-log-tabexpand.sh.
diff --git a/Documentation/git-check-ignore.txt b/Documentation/git-check-ignore.txt index 2892799e32f..69d12fd22e1 100644 --- a/Documentation/git-check-ignore.txt +++ b/Documentation/git-check-ignore.txt @@ -50,10 +50,8 @@ linkgit:gitignore[5]. with a NUL character instead of a linefeed character. -n, --non-matching:: - Show given paths which don't match any pattern. This only - makes sense when `--verbose` is enabled, otherwise it would - not be possible to distinguish between paths which match a - pattern and those which don't. + Only show given paths which don't match any pattern. If `--verbose` is + enabled, show both paths that match a pattern and those which don't. --no-index:: Don't look in the index when undertaking the checks. This can diff --git a/builtin/check-ignore.c b/builtin/check-ignore.c index 21912569650..10e8bba1057 100644 --- a/builtin/check-ignore.c +++ b/builtin/check-ignore.c @@ -113,8 +113,17 @@ static int check_ignore(struct dir_struct *dir, pattern->flags & PATTERN_FLAG_NEGATIVE) pattern = NULL; } - if (!quiet && (pattern || show_non_matching)) - output_pattern(pathspec.items[i].original, pattern); + if (!quiet) { + if (verbose) { + if (show_non_matching || pattern) + output_pattern(pathspec.items[i].original, pattern); + } else { + if (show_non_matching && !pattern) + output_pattern(pathspec.items[i].original, pattern); + if (!show_non_matching && pattern) + output_pattern(pathspec.items[i].original, pattern); + } + } if (pattern) num_ignored++; } @@ -175,8 +184,6 @@ int cmd_check_ignore(int argc, const char **argv, const char *prefix) if (verbose) die(_("cannot have both --quiet and --verbose")); } - if (show_non_matching && !verbose) - die(_("--non-matching is only valid with --verbose")); /* read_cache() is only necessary so we can watch out for submodules. */ if (!no_index && read_cache() < 0) diff --git a/t/t0008-ignores.sh b/t/t0008-ignores.sh index 5575dade8ee..2ea178797bc 100755 --- a/t/t0008-ignores.sh +++ b/t/t0008-ignores.sh @@ -90,7 +90,8 @@ test_check_ignore () { # 1. with -q / --quiet # 2. with default verbosity # 3. with -v / --verbose -# 4. with -v / --verbose, *and* -n / --non-matching +# 4. with -n / --non-matching +# 5. with -v / --verbose, *and* -n / --non-matching # # expecting success each time. Takes advantage of the fact that # check-ignore --verbose output is the same as normal output except @@ -102,7 +103,7 @@ test_check_ignore () { # Arguments: # - (optional) prereqs for this test, e.g. 'SYMLINKS' # - test name -# - output to expect from the fourth verbosity mode (the output +# - output to expect from the fifth verbosity mode (the output # from the other verbosity modes is automatically inferred # from this value) # - code to run (should invoke test_check_ignore) @@ -124,6 +125,7 @@ test_expect_success_multiple () { expect_verbose=$( echo "$expect_all" | grep -v '^:: ' ) expect=$( echo "$expect_verbose" | sed -e 's/.* //' ) + expect_non_matching=$( echo "$expect_all" | grep '^:: ' | sed -e 's/.* //') test_expect_success $prereq "$testname${no_index_opt:+ with $no_index_opt}" ' expect "$expect" && @@ -144,6 +146,17 @@ test_expect_success_multiple () { quiet_opt= fi + for non_matching_opt in '-n' '--non-matching' + do + test_code=" + expect '$expect_non_matching' && + $code + " + opts="${no_index_opt:+$no_index_opt }$non_matching_opt" + test_expect_success $prereq "$testname${opts:+ with $opts}" "$test_code" + done + non_matching_opt= + for verbose_opt in '-v' '--verbose' do for non_matching_opt in '' '-n' '--non-matching'