bug-guix
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

bug#28446: address@hidden should select 4.1.x, not 4.13.x


From: Mark H Weaver
Subject: bug#28446: address@hidden should select 4.1.x, not 4.13.x
Date: Tue, 26 Sep 2017 00:26:15 -0400
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/25.3 (gnu/linux)

Hi Ricardo,

Ricardo Wurmus <address@hidden> writes:

>> Currently, the package specification "address@hidden" selects version
>> 4.13.  It should instead select version 4.1.
>
> We consider everthing following the address@hidden a version string prefix.  
> Since
> versions are arbitrary strings “4.1” is considered a valid prefix of
> “4.13”.  If a user supplied the version string “4.1.” they would get the
> appropriate package.
>
> The current implementation sorts all matches in decreasing version order
> and picks the package with the highest version.  This is implemented in
> (gnu packages) with “%find-packages”, “find-best-packages-by-name”, and
> “find-packages-by-name”.
>
> Should we try to make the code understand version strings better and
> compare substrings of the version string?  We could fall back to using
> “string-prefix?” when the substring is not a number.

One possible approach would be as follows: if the last character of the
specification is a digit, then priority should be given to versions that
have a non-digit (or end-of-string) after the last matching character.

Put another way, if the provided specification {SPEC} ends with a digit,
then we should first search based on the following regexp:

  ^{SPEC}[^0-9]

if no matches are found, then fall back to:

  ^{SPEC}

This would gracefully handle the common case where groups of adjacent
digits should be considered as single numbers, while still allowing
prefixes of other numeric components such as dates or commit ids.

I admit that this proposal lacks elegance.  So far I have no better
ideas.

      Mark





reply via email to

[Prev in Thread] Current Thread [Next in Thread]