Git Mailing List Archive on lore.kernel.org
 help / color / Atom feed
From: Patrick Steinhardt <ps@pks.im>
To: Junio C Hamano <gitster@pobox.com>
Cc: git <git@vger.kernel.org>, Christian Couder <christian.couder@gmail.com>
Subject: Re: [PATCH 4/9] update-ref: organize commands in an array
Date: Mon, 30 Mar 2020 10:05:51 +0200
Message-ID: <20200330080551.GA186019@ncase.pks.im> (raw)
In-Reply-To: <xmqqh7y9fq29.fsf@gitster.c.googlers.com>


[-- Attachment #1: Type: text/plain, Size: 4976 bytes --]

On Fri, Mar 27, 2020 at 02:25:34PM -0700, Junio C Hamano wrote:
> Patrick Steinhardt <ps@pks.im> writes:
> 
> > +static const struct parse_cmd {
> > +	const char *prefix;
> > +	const char *(*fn)(struct ref_transaction *, struct strbuf *, const char *);
> > +} commands[] = {
> 
> Do not call an array the represents a set of THINGs "type things[]";
> instead call it "type thing[]", so that the 0th thing can be
> referred to as thing[0], not things[0].
> 
> One exception is when the set as a whole is referred to more often
> than individual element of an array, in which case "things" (without
> the [index]) becomes a sensible way to refer to the set.
> 
> > +	{ "update", parse_cmd_update },
> > +	{ "create", parse_cmd_create },
> > +	{ "delete", parse_cmd_delete },
> > +	{ "verify", parse_cmd_verify },
> > +	{ "option", parse_cmd_option },
> > +};
> > +
> >  static void update_refs_stdin(struct ref_transaction *transaction)
> >  {
> >  	struct strbuf input = STRBUF_INIT;
> >  	const char *next;
> > +	int i;
> >  
> >  	if (strbuf_read(&input, 0, 1000) < 0)
> >  		die_errno("could not read from stdin");
> >  	next = input.buf;
> >  	/* Read each line dispatch its command */
> >  	while (next < input.buf + input.len) {
> > +		const struct parse_cmd *cmd = NULL;
> > +
> >  		if (*next == line_termination)
> >  			die("empty command in input");
> >  		else if (isspace(*next))
> >  			die("whitespace before command: %s", next);
> > -		else if (skip_prefix(next, "update ", &next))
> > -			next = parse_cmd_update(transaction, &input, next);
> > -		else if (skip_prefix(next, "create ", &next))
> > -			next = parse_cmd_create(transaction, &input, next);
> > -		else if (skip_prefix(next, "delete ", &next))
> > -			next = parse_cmd_delete(transaction, &input, next);
> > -		else if (skip_prefix(next, "verify ", &next))
> > -			next = parse_cmd_verify(transaction, &input, next);
> > -		else if (skip_prefix(next, "option ", &next))
> > -			next = parse_cmd_option(&input, next);
> > -		else
> > +
> > +		for (i = 0; i < ARRAY_SIZE(commands); i++) {
> > +			if (!skip_prefix(next, commands[i].prefix , &next))
> > +				continue;
> > +			cmd = &commands[i];
> > +			break;
> > +		}
> 
> The only reason why you had to sprinkle
> 
> 	if (!skip_prefix(next, " ", &next))
> 		die(_("%s: missing space after command"), cmd);
> 
> all over the place is because the table lacks the trailing SP (which
> makes sense---after all, you are making a table of commands).  In
> other words, it's not like some command called from this dispatcher
> would require " " after the command name and some others would not.

> So why not avoid touching the parse_cmd_<cmd>() at all (except for
> the "option" thing that now needs to take the transaction object for
> uniformity), and then verify the presence of " " here, perhaps like
> this:
> 
> 	for (i = 0; i < ARRAY_SIZE(command); i++) {
> 		const char *eoc;
> 		if (!skip_prefix(next, commands[i].prefix, &eoc) ||
> 		    *eoc != ' ')
> 			continue;
> 		cmd = &command[i];
>                 next = eoc;
> 		break;
> 	}

The reason why I moved those `skip_prefix` calls into each of the
respective commands is that this patch series introduces calls that do
not accept a trailing space at all. Thus we cannot handle the space
generically here, as that would was soon as we introduce the set of new
commands.

Initially I just had a trailing " " in the `command` array as you hint
at, but it felt a bit magical to me and might make readers wonder why
some commands are spelt "update " and why some are spelt "commit",
without a trailing space.

> Note that you cannot reuse &next here to future-proof the code;
> otherwise, you wouldn't be able to add a new command, e.g. "options",
> that sits next to the existing command "option", in the future.
> 
> > +		if (!cmd)
> >  			die("unknown command: %s", next);
> >  
> > +		if (input.buf[strlen(cmd->prefix)] != line_termination &&
> > +		    input.buf[strlen(cmd->prefix)] != '\0' &&
> > +		    input.buf[strlen(cmd->prefix)] != ' ')
> > +			die("%s: no separator after command", cmd->prefix);
> 
> This part of your version does not make sense to me.  If the input
> line began with "update" with some separator that is not a SP, the
> original would not have matched.  But with this code, "update\0"
> would pass this code and cause cmd->fn() to be called, only to get
> the input rejected, because next is not pointing to SP.  So why do
> you even need this if statement?

I hope this makes more sense with the above background of commands
without trailing space. It would probably make sense to move this into
the command-matching loop though to be able to discern "option" and a
potentially new "options" command in the future. Instead of dying, we'd
then have to `continue` the loop and check whether any of the remaining
commands matches.

Patrick

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

  reply index

Thread overview: 48+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-25  9:53 [PATCH 0/9] Support for transactions in `git-update-ref --stdin` Patrick Steinhardt
2020-03-25  9:53 ` [PATCH 1/9] refs: fix segfault when aborting empty transaction Patrick Steinhardt
2020-03-27 19:59   ` Junio C Hamano
2020-03-25  9:53 ` [PATCH 2/9] git-update-ref.txt: add missing word Patrick Steinhardt
2020-03-25  9:53 ` [PATCH 3/9] strbuf: provide function to append whole lines Patrick Steinhardt
2020-03-27 21:04   ` Junio C Hamano
2020-03-30 13:25     ` Patrick Steinhardt
2020-03-30 17:12       ` Junio C Hamano
2020-03-25  9:53 ` [PATCH 4/9] update-ref: organize commands in an array Patrick Steinhardt
2020-03-27 21:25   ` Junio C Hamano
2020-03-30  8:05     ` Patrick Steinhardt [this message]
2020-03-30 16:55       ` Junio C Hamano
2020-03-30 17:37         ` Patrick Steinhardt
2020-03-25  9:54 ` [PATCH 5/9] update-ref: drop unused argument for `parse_refname` Patrick Steinhardt
2020-03-25  9:54 ` [PATCH 6/9] update-ref: pass end pointer instead of strbuf Patrick Steinhardt
2020-03-25  9:54 ` [PATCH 7/9] update-ref: move transaction handling into `update_refs_stdin()` Patrick Steinhardt
2020-03-27 21:44   ` Junio C Hamano
2020-03-25  9:54 ` [PATCH 8/9] update-ref: read commands in a line-wise fashion Patrick Steinhardt
2020-03-27 21:58   ` Junio C Hamano
2020-03-30  8:11     ` Patrick Steinhardt
2020-03-30 17:39       ` Junio C Hamano
2020-03-25  9:54 ` [PATCH 9/9] update-ref: implement interactive transaction handling Patrick Steinhardt
2020-03-27 22:00   ` Junio C Hamano
2020-03-30 13:46 ` [PATCH v2 0/9] Support for transactions in `git-update-ref --stdin` Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 1/9] refs: fix segfault when aborting empty transaction Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 2/9] git-update-ref.txt: add missing word Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 3/9] strbuf: provide function to append whole lines Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 4/9] update-ref: organize commands in an array Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 5/9] update-ref: drop unused argument for `parse_refname` Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 6/9] update-ref: pass end pointer instead of strbuf Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 7/9] update-ref: move transaction handling into `update_refs_stdin()` Patrick Steinhardt
2020-03-30 13:46   ` [PATCH v2 8/9] update-ref: read commands in a line-wise fashion Patrick Steinhardt
2020-03-30 13:47   ` [PATCH v2 9/9] update-ref: implement interactive transaction handling Patrick Steinhardt
2020-04-02  7:09 ` [PATCH v3 0/9] Support for transactions in `git-update-ref --stdin` Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 1/9] refs: fix segfault when aborting empty transaction Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 2/9] git-update-ref.txt: add missing word Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 3/9] strbuf: provide function to append whole lines Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 4/9] update-ref: organize commands in an array Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 5/9] update-ref: drop unused argument for `parse_refname` Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 6/9] update-ref: pass end pointer instead of strbuf Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 7/9] update-ref: move transaction handling into `update_refs_stdin()` Patrick Steinhardt
2020-04-02  7:09   ` [PATCH v3 8/9] update-ref: read commands in a line-wise fashion Patrick Steinhardt
2020-04-02  7:10   ` [PATCH v3 9/9] update-ref: implement interactive transaction handling Patrick Steinhardt
2020-04-03 13:40     ` Phillip Wood
2020-04-03 16:51       ` Patrick Steinhardt
2020-04-03 17:33       ` Junio C Hamano
2020-04-03 17:35         ` Junio C Hamano
2020-04-06  7:10         ` Patrick Steinhardt

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200330080551.GA186019@ncase.pks.im \
    --to=ps@pks.im \
    --cc=christian.couder@gmail.com \
    --cc=git@vger.kernel.org \
    --cc=gitster@pobox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Git Mailing List Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/git/0 git/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 git git/ https://lore.kernel.org/git \
		git@vger.kernel.org
	public-inbox-index git

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.git


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git