Re: [PATCH/RFC] blame: accept multiple -L ranges

From: Thomas Rast <trast@inf.ethz.ch>
Date: Jul 22, 2013, 19:19 UTC
Message-ID: <87bo5uqu6i.fsf@linux-k42r.v.cablecom.net>
In-Reply-To: <7vr4eqcxub.fsf@alter.siamese.dyndns.org>

Junio C Hamano <gitster@pobox.com> writes:

Show 36 quoted lines

> Thomas Rast <trast@inf.ethz.ch> writes:
>
>> Eric Sunshine <sunshine@sunshineco.com> writes:
>>
>>> The proposal currently is only for "-L /RE/,whatever" to behave in a
>>> relative fashion, beginning the search at the end of the last range
>>> specified via -L (or line 1 if there is no previous -L).
>>>
>>> Would it also make sense to support "-L +N,whatever" as relative to
>>> the end of the last range specified via -L (or 1 if none).
>>
>> Sounds reasonable.
>>
>> I'm still not sure I am super-happy with /RE/ always being relative,
>> though I see Junio's problem space as something worth solving.  How does
>> it interact with -L:RE?  Do you now have to know in what order the
>> functions appear in the source to correctly specify -L:foo -L:bar or
>> similarly, -L/foo/,/^}/ -L/bar/,/^}/?  What if we supported +/RE/ as the
>> relative version?
>
> Two gripes I have are:
>
>  (1) That sounds like making common things more cumbersome to ask.
>
>  (2) In "-L /RE1/,/RE2/", you do not have to say +/RE2/ (and you
>      shouldn't have to).  /RE3/ without any magic that starts
>      searching after the last match in "-L /RE1/,/RE2/ -L /RE3/,+4"
>      feels a lot more consistent than requiring a prefix plus.
>
> I am OK if you made /RE/, which starts searching immediately after
> the last match, wrap around and continue the search at the beginning
> upon finding nothing through the end of the file (and make sure you
> stop if you passed the last match again).  That would solve your "I
> have two functions, and I can give them in any order", while keeping
> the consistency with "In /RE1/,/RE2/, the latter is already relative
> to the former".

Dunno. It still doesn't really solve the order-of-L problem if there are multiple matches.

I can't really say how they fare against each other. I have a bad feeling about the consistency of what results, but as you say, doing it the way I suggested isn't very consistent either. Perhaps in retrospect even -L/foo/,/bar/ should have required the + in +/bar/ to make it relative.

I'll just leave it at that and let you decide what to do (presumably go ahead as you already outlined). I've never actually ever used multiple -L in the same log/blame invocation, anyway. I just think that when it comes to it, I'll have to read the manpage before I try...

-- 
Thomas Rast
trast@{inf,student}.ethz.ch