[Differential] [Commented On] D59: sparse: override dirstate.walk() instead of dirstate._ignore

[durham (Durham Goode)]

durham added a comment.


  My only concern would be if tools ever use hg status -A to get the full state of the working copy (like what files exist). But I think that's pretty minor.
  
  Do we know if the ignore wrapper was ever accessed outside of the walk?  It might be worth hacking in a temporary message that prints if the ignorewrapper is called while walk is not in the traceback, then run the tests to see if it's ever hit.
  
  Since this logic makes it the caller's responsibility to pass the right matcher to walk(), we may want to find a call pattern that doesn't require every caller to read the sparse matcher and union it with their own matcher.  Perhaps by adding the sparse matcher by default (inside walk), unless explicitly requested otherwise.  Or by having commands take the user specified patterns and passing them to a workingcopymatcher() or something that will automatically take into account the sparse checkout (you could have a storematcher() as well for taking into account narrowness too).

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D59

EMAIL PREFERENCES
  https://phab.mercurial-scm.org/settings/panel/emailpreferences/

To: martinvonz
Cc: durham, dsp, mercurial-devel