Re: [PATCH 2/2] git.c: binary-search builtin commands

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 15, 2011, 20:12 UTC
Message-ID: <7v4o85njjz.fsf@alter.siamese.dyndns.org>
In-Reply-To: <1297739344-28961-2-git-send-email-pclouds@gmail.com>

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

Show 12 quoted lines

> An obvious implication of this patch: commands array must be in correct
> order.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>   2011/2/15 Junio C Hamano <gitster@pobox.com>:
>   > I did this myself the other day, as I think it simply is a good project
>   > hygiene.  If this were 1/2 of a series followed by 2/2 that runs binary
>   > search in the table, that would make it make more sense ;-)
>   
>   I did think the array was binary-searched and nearly claimed "git-stage
>   won't work because it's in wrong order".

Heh, that "binary search" was a tongue-in-cheek comment. I am sorry that you took it too seriously.

>   This patch won't give any performance gain, but it would force
>   people to keep the array in order :-)

That is exactly why I discarded what I did the other day. Without an active mechanism to force the orderedness, such a change simply introduces a downside of letting a mistake go unnoticed, without any real upside (as you measured and saw no performance gain).

A better project hygine is a good thing to aim for, and I would imagine that you could add "--verify-builtin-command-table" as an unadvertised option to "git" wrapper, and make t/t0000-basic.sh call it to minimize the downside risk. But without such an active measure to prevent mistakes, we would be relying on somebody getting caught on a ticking bomb and reporting it, which is not a good tradeoff between risk and reward.