From: Eric Sunshine Date: Tue, 02 Jul 2013 19:04:17 GMT Subject: Re: [PATCH/RFC 1/4] contrib: add git-contacts helper Message-ID: In-Reply-To: <7vk3l895in.fsf@alter.siamese.dyndns.org> On Tue, Jul 2, 2013 at 2:32 PM, Junio C Hamano wrote: > Eric Sunshine writes: > >>> The author name and email can be grabbed from the "blame" output >>> without doing this (and the result may be more robust), but you >>> would need to read from the log message anyway, so I think this is >>> OK. >>> >>> Note that the names and emails in blame output are sanitized via the >>> mailmap mechanism, but "cat-file commit" will certainly not be. >> >> Thanks for pointing this out. Grabbing the author name and email from >> git-blame output does seem like a better approach. > > Well, that was not what I was suggesting. Reading from blame output > will map only the author/committer names/addresses, and you have two > choices: > > (1) if you read author/committer from blame output and other names > from the commit object without applying mailmap, the same > person can appear under different names, from his authorship > (mapped name) and from his name on footers (unmapped), which > would be inconsistent. By reading from "author" and > "committer" header lines in the commit object, you will be at > least consistently using unmapped names. > > (2) if you want to apply mailmap to names you collect by reading > the footer, you will write the mapping logic yourself anyway, > and at that point, passing the names you collect by reading the > "author" and "committer" header lines in the commit object to > the same logic will give you mapped names. At that point, you > do not gain much by reading names from the blame output. > > So in either case, you would be better off not reading the > author/committer from blame output, as long as you need to pick up > other names from the commit object payload. Thanks for the clarification.