archive.getPublishedSources matches package_name as substring

[Martin Pool]

It looks like getPublishedSources matches source_name as a substring by default, so searching for 'under' returns 'thunderbird' amongst others.

This seems at least a bit weird. Apparently there is an exact_match option.

[Julian Edwards (julian-edwards)]

exact_match defaults to False. I don't think this can realistically be changed now without affecting all the users of this.

[Julian Edwards (julian-edwards)]

Actually let me re-phrase that, how would you change this? I'm happy to accept any reasonable suggestion :)

[Martin Pool (mbp)]

2010/1/26 Julian Edwards <email address hidden>:
> exact_match defaults to False.  I don't think this can realistically be
> changed now without affecting all the users of this.

Is inexact matching a general pattern in Launchpad APIs? It seems pretty weird.

--
Martin <http://launchpad.net/~mbp/>

[Julian Edwards (julian-edwards)]

On Tuesday 26 January 2010 17:41:25 Martin Pool wrote:
> 2010/1/26 Julian Edwards <email address hidden>:
> > exact_match defaults to False. I don't think this can realistically be
> > changed now without affecting all the users of this.
>
> Is inexact matching a general pattern in Launchpad APIs? It seems
> pretty weird.
>

I don't think it is. Unfortunately this API was exposed as a complete match
of what the internal method is, without really thinking about it :(

[Martin Pool (mbp)]

2010/1/26 Julian Edwards <email address hidden>:
> On Tuesday 26 January 2010 17:41:25 Martin Pool wrote:
>> 2010/1/26 Julian Edwards <email address hidden>:
>> > exact_match defaults to False.  I don't think this can realistically be
>> > changed now without affecting all the users of this.
>>
>> Is inexact matching a general pattern in Launchpad APIs?  It seems
>> pretty weird.
>
> I don't think it is.  Unfortunately this API was exposed as a complete match
> of what the internal method is, without really thinking about it :(

So I would hazard a guess that if anyone is using this api, they're
probably getting confusing results in the same way that we were, and
if they really want substring matching they are probably saying so.

However, perhaps we could at least add a big warning about this to the
apidoc? I suppose getPublishedBinaries etc would be similar.

--
Martin <http://launchpad.net/~mbp/>

[Julian Edwards (julian-edwards)]

Well what we could do is change the default and just suck it up. I know the
two main users of this method and it would be easy to warn them. A message to
-users should cover the rest.

I agree that it's a bit crack.