[Qa-debsources] DB modification for cloc plugin

[Orestis Ioannou]

Heya

On 04/13/2016 05:18 PM, Stefano Zacchiroli wrote:
> On Wed, Apr 13, 2016 at 10:24:20AM +0200, Matthieu Caneill wrote:
>> Thanks for working on this!
> 
> Hear, hear.
> 
>> Since it's about cloc, and not likely to change (other kind of meta
>> information should be added in the future directly in the Files
>> table), I would suggest file_cloc.
> 
> Or, if you feel funny enough, we can use cloccounts, which is just one
> character away from sloccounts, but is incidentally also very
> semantically appropriate.  I think with indexes doing the aggregate by
> package would actually be really fast, and could be done on the fly,
> rather than having a separate dedicated table. It's worth being tested,
> but I'm actually pretty confident it'd work fine.


cloccounts seems good to me :)
Do we need a package_id in the table for the indexes or this is not what
you meant?


> 
> Yeah, definitely.
> 
>>> CREATE TABLE file_metas (
>>>   id SERIAL NOT NULL,
>>>   file_id BIGINT NOT NULL,
>>>   code Integer NOT NULL,
>>>   comment Integer,
>>>   language languages NOT NULL,
>>>   CONSTRAINT file_metas_file_id_fkey
>>>     FOREIGN KEY (file_id) REFERENCES files(id)
>>>     ON DELETE CASCADE,
>>>   PRIMARY KEY (id)
>>> );
>>
>> * We should create a UNIQUE constraint for (file_id, language);
> 
> ... and drop id, yes (unless you can think of specific reasons why we
> would event want to point to a specific line count entry).
> 
> Question: can cloc actually return multiple languages for the same file?
> 
> (The above constraint seems to conservatively assume that it could,
> which is a nice assumption for the DB, but I'm curious anyhow)
> 


So the thing is that i thought it could return many languages per file.
For example a html file with some php in it etc. But i tested this today
and its not working like tihs. But at the beginnning this is why i
thought of an id because the unique constraint wouldn't work.

I ll ask the author about this and modify accordingly.

>> * I would append the suffix _count to code and comment;
>> * Why drop the blank lines count of cloc? FWIW, once we have it it's
>>   almost free to store.
> 
> *nod*
> 

ACK. Thanks again both of you for the review :)

Orestis

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/qa-debsources/attachments/20160414/e2d938de/attachment.sig>