Re: [PATCH] use skip_prefix() to avoid more magic numbers

From: Junio C Hamano <gitster@pobox.com>
Date: Oct 7, 2014, 18:23 UTC
Message-ID: <xmqqd2a3g2mf.fsf@gitster.dls.corp.google.com>
In-Reply-To: <5430427A.5080800@web.de>

René Scharfe <l.s.r@web.de> writes:

Show 24 quoted lines

> @@ -335,20 +337,18 @@ static int append_ref(const char *refname, const unsigned char *sha1, int flags,
>  	static struct {
>  		int kind;
>  		const char *prefix;
> -		int pfxlen;
>  	} ref_kind[] = {
> -		{ REF_LOCAL_BRANCH, "refs/heads/", 11 },
> -		{ REF_REMOTE_BRANCH, "refs/remotes/", 13 },
> +		{ REF_LOCAL_BRANCH, "refs/heads/" },
> +		{ REF_REMOTE_BRANCH, "refs/remotes/" },
>  	};
>  
>  	/* Detect kind */
>  	for (i = 0; i < ARRAY_SIZE(ref_kind); i++) {
>  		prefix = ref_kind[i].prefix;
> -		if (strncmp(refname, prefix, ref_kind[i].pfxlen))
> -			continue;
> -		kind = ref_kind[i].kind;
> -		refname += ref_kind[i].pfxlen;
> -		break;
> +		if (skip_prefix(refname, prefix, &refname)) {
> +			kind = ref_kind[i].kind;
> +			break;
> +		}

This certainly makes it easier to read.

I suspect that the original was done as a (possibly premature) optimization to avoid having to do strlen(3) on a variable that points at constant strings for each and every ref we iterate with for_each_rawref(), and it is somewhat sad to see it lost because skip_prefix() assumes that the caller never knows the length of the prefix, though.

Thanks.