Re: [Gnu-arch-users] names -> tagline method transition

[Michael Teichgräber]

Tom Lord <address@hidden> writes:

>     > the transition will recognize most files as deleted+added. Would it
>     > have drawbacks to treat ids beginning with "?_" and "?" as the same
>     > when calculating changes?
>
> It wouldn't be an upwards compatible change.

The =TODO file still contains a section about =tagging-method
generalization, where names, explicit, and tagline will be constructed
by keywords such as external-tag, embedded-tag. If I didn't miss
something this is not yet fully implemented (apart from
untagged-source).

Would this mechanism be upwards compatible? If `names' and `tagline'
are just short-hands for certain combinations of values, then perhaps
`untagged-source source' in a setting resembling former `tagline'
would result in the same id for a simple file as for a setting that
equals former `names'. Otherwise there would have to be some hidden
mechanism that generates "?_./name" in one and "?./name" in the other
case, or that treats a specified `names' different than a simple
`untagged-source source'.


> New users often ask for this feature.  They get started quickly by
> using names or explicit method, then want to transition, and get
> annoyed at the discontinuity.  I don't think that that's a good reason
> to add the feature.

That project was actually one of my first projects managed with arch,
when `names' was still the default (if I remember right) and I perhaps
got started too quickly. After not working on it for some time, and
now wanting to transition, I made the experience you described in the
last part of that sentence ;-)

> A recent trend is more people looking into tracking external projects
> that aren't natively handled in arch.   Would this feature help a lot
> with that?   I think it might.   Not so much for
> names->{explicit,tagline} but for (file-by-file) explicit->embedded.

If `tagline' + `untagged-source source' nearly provides at least the
same features as `names' (with some accidentally detected embedded
tags perhaps), wouldn't it make sense to discontinue `names' in some
way? Some thoughts:

* One could begin to hide it from the help messages and documentation,
  so that projects currently using it may proceed, but new users won't
  try "tla id-tagging-method names".

* Or, if someone enters "tla i-t-m names", tla could actually insert
  `tagline' + `untagged-source source' into =tagging-method (or just
  the second directive, as explained in the next paragraph).

  To provide that upwards compatibility mentioned in the beginning, if
  tla detects a `names' directive in =tagging-method, it could work as
  usual, i.e. generate "?./name" ids (this would only apply to "old"
  projects).

  In contrast, with `names' missing from =tagging-method and no
  external-tag or embedded-tag, but `untagged-source source'
  specified, it would generate "?_./name" (and "A_..." for files under
  {arch}).

  I wonder if it would be tricky to implement this.
  `arch_names_id_tagging' doesn't occur at many places, one would have
  to code slightly different sections for an additional (e.g.)
  `arch_newnames_id_tagging' that will produce untagged file's ids
  more consistent with those in explicit|tagline trees. Also `srcfind'
  and `id' would have to be adapted slightly. Even in case
  external-tag and embedded-tag don't get implemented soon, that seems
  achievable to me without making things more complicated.


-- 
Michael