Re: [Monotone-devel] Re: write access to my public server (was Re: [PATCH] add xemacs ...)

[Christof Petig]

Derek Scherger schrieb:
> graydon hoare wrote:
>
> > unfortunately revert already takes filenames as an argument. it might 
> > make sense to make a manifest and file variant of revert. njs has 
> > proposed that maybe commands should just get an 'f' or 'm' prefix, 
> > i.e. mrevert or frevert, mlog or flog, mdiff or fdiff. what do you 
> > think about that convention?

This case looks difficult. Perhaps the x: convention might be a good 
idea here, too. I would prefer less commands to remember - or leave the 
choice up to taste (perhaps a bad idea)

  monotone revert xyz
=> if there exists a file named xyz -> file
=> if it might be an id look for a matching manifest
  monotone frevert xyz    # must be a file
  monotone revert f:xyz   # file
  monotone mrevert xyz    # manifest ID
  monotone revert i:xyz   # manifest ID

> I've been wondering whether commands like cat and complete really need a 
> file/manifest type at all. Presumably the version codes are unique and 
> only live as either a file or a manifest. What about this:
>
> $ monotone complete 5d
> 5d2e96646f7be38c0aceacf9331ab687e8130f93 (manifest)
> $ monotone complete 45
> 45229a1c4745275390cee1a1a60433db5e82578e (file)
>
> Where monotone tells the user what type of object the manifest id refers 
> to?

I second this view. Make monotone find out whether you mean manifest or 
file. If it's ambiguous we can prepend it with m:(or i: for consistency) 
or f: (if these are not taken, this is modelled after the i:ID, d:date, 
r:tag convention used by diff (? from memory ?))

> As far as f... and m... command variants go, I think I'd favour a 
> smaller number of commands with some options. Perhaps monotone diff -f 
> for files and monotone diff -m for manifests would be good in cases 
> where the above idea doesn't work.

Can we please use less strange options. I would like to leave open the 
option of somewhat cvs option compatibility (but without its strange two 
independant sets (before and after the action)):
-m message on checkin (not commandline possibility with monotone, yet)
-d ==> --db

I like the i:/d: ... specifiers more (since they are only needed when 
ambiguous)

   Christof