Re: [RFC 1/9] lib/string: Add function to trim duplicate WS

From: Tony Asleson <tasleson@redhat.com>
To: Matthew Wilcox <willy@infradead.org>
Cc: linux-scsi@vger.kernel.org, linux-block@vger.kernel.org,
	linux-fsdevel@vger.kernel.org
Subject: Re: [RFC 1/9] lib/string: Add function to trim duplicate WS
Date: Thu, 2 Jan 2020 16:52:19 -0600	[thread overview]
Message-ID: <8392b726-fa55-baa4-6913-5ca0e4fa46a7@redhat.com> (raw)
In-Reply-To: <20191223232824.GB31820@bombadil.infradead.org>

On 12/23/19 5:28 PM, Matthew Wilcox wrote:
> On Mon, Dec 23, 2019 at 04:55:50PM -0600, Tony Asleson wrote:
>> +/**
>> + * Removes leading and trailing whitespace and removes duplicate
>> + * adjacent whitespace in a string, modifies string in place.
>> + * @s The %NUL-terminated string to have spaces removed
>> + * Returns the new length
>> + */
> 
> This isn't good kernel-doc.  See Documentation/doc-guide/kernel-doc.rst
> Compile with W=1 to get the format checked.

Indeed, I'll correct it.

>> +size_t strim_dupe(char *s)
>> +{
>> +	size_t ret = 0;
>> +	char *w = s;
>> +	char *p;
>> +
>> +	/*
>> +	 * This will remove all leading and duplicate adjacent, but leave
>> +	 * 1 space at the end if one or more are present.
>> +	 */
>> +	for (p = s; *p != '\0'; ++p) {
>> +		if (!isspace(*p) || (p != s && !isspace(*(p - 1)))) {
>> +			*w = *p;
>> +			++w;
>> +			ret += 1;
>> +		}
>> +	}
> 
> I'd be tempted to do ...
> 
> 	size_t ret = 0;
> 	char *w = s;
> 	bool last_space = false;
> 
> 	do {
> 		bool this_space = isspace(*s);
> 
> 		if (!this_space || !last_space) {
> 			*w++ = *s;
> 			ret++;
> 		}
> 		s++;
> 		last_space = this_space;
> 	} while (s[-1] != '\0');

That leaves a starting and trailing WS, how about something like this?

size_t strim_dupe(char *s)
{
	size_t ret = 0;
	char *w = s;
	bool last_space = false;

	do {
		bool this_space = isspace(*s);
		if (!this_space || (!last_space && ret)) {
			*w++ = *s;
			ret++;
		}
		s++;
		last_space = this_space;
	} while (s[-1] != '\0');

	if (ret > 1 && isspace(w[-2])) {
		w[-2] = '\0';
		ret--;
	}

	ret--;
	return ret;
}

Thanks
-Tony