[PATCHv7 0/5] pathspec magic extension to search for attributes

This goes on top of origin/jc/attr, (8195c5cde01, SQUASH???, 2016-05-17 14:08:50)

Patches 1 and 2 are small fixes, which could go independently as well. (1 is new compared to v6)

Patches 3 and 4 are refactoring pathspec.c a little.

The last patch is the most exciting patch as it adds the new feature. Quoting from the documentation update:

        attr;;
                Additionally to matching the pathspec, the path must have the
                attribute as specified. The syntax for specifying the required
                attributes is "`attr: [mode] <attribute name> [=value]`"
        +
        Attributes can have 4 states (Set, Unset, Set to a value, unspecified) and
        you can query each attribute for certain states. The "`[mode]`" is a special
        character to indicate which attribute states are looked for. The following
        modes are available:
        
        - "`+`" the attribute must be set
        - "`-`" the attribute must be unset
        - "`~`" the attribute must be unspecified
        - "`?`" the attribute must not be unspecified, i.e. set, unset or value matches
        - an empty "`[mode]`" matches if the attribute is set or has a value
        - an empty "`[mode]`" combined with "`[=value]`" matches if the attribute has
          the given value.
        +

While I followed Junios suggestion of the data layout in pathspec.h, I am not quite happy with it yet as I think it will not be easy in the future to enhance search for specific attribute values. Quoting from the tests:

        test_expect_success 'check label with 2 labels' '
                cat <<EOF >expect &&
        fileAB     # has both labels
        sub/fileAB # has both labels
        EOF
                git ls-files ":(attr:labelA labelB)" >actual &&
                test_cmp expect actual &&
                git ls-files ":(attr:labelA,attr:labelB)" >actual &&
                test_cmp expect actual
        '

You can see that two notations are possible to narrow the search down when asking for 2 attributes to be set. Either you can do 2 "attr:" magic strings or you can have one "attr:" and then a space separated list of attributes.

The problem arises once you'd want to have a better search for values, as currently only ":(attr:label=value) is supported with "value" being the exact match, e.g. a path having more than just value would not match as it is not exact. E.g. in git.git we have

        $ cat .gitattributes 
        * whitespace=!indent,trail,space
        *.[ch] whitespace=indent,trail,space
        *.sh whitespace=indent,trail,space
        $ git ls-files ":(attr:whitespace=indent)"
        $ # no match!

So instead of matching exactly we'd want to just look for the value being one part of the comma separated list as setup in .gitattributes. However searching for more than one value will be complicated, as we'd then need to escape commas. My vision was to use white spaces instead, but they are now taken to separate different attributes.

So in the future I would imagine to have escaping like 

        git ls-files ":(attr:label=value\,value2 labelA=value3,exclude)
        
or yet another separator like

        git ls-files ":(attr:label=value1;value2 labelA=value3,exclude)

Thanks, Stefan

        
Stefan Beller (5):
  string list: improve comment
  Documentation: fix a typo
  pathspec: move long magic parsing out of prefix_pathspec
  pathspec: move prefix check out of the inner loop
  pathspec: allow querying for attributes

 Documentation/gitattributes.txt    |   2 +-
 Documentation/glossary-content.txt |  19 ++++
 attr.c                             |   2 +-
 attr.h                             |   2 +
 dir.c                              |  46 ++++++++++
 pathspec.c                         | 182 +++++++++++++++++++++++++++++--------
 pathspec.h                         |  16 ++++
 string-list.h                      |   2 +-
 t/t6134-pathspec-with-labels.sh    | 177 ++++++++++++++++++++++++++++++++++++
 9 files changed, 407 insertions(+), 41 deletions(-)
 create mode 100755 t/t6134-pathspec-with-labels.sh