git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v2 02/19] parse-options-cb: implement parse_opt_pass_argv_array()

From
Junio C Hamano <gitster@pobox.com>
Date
Jun 10, 2015, 08:03 UTC
Message-ID
<xmqqy4jst2fp.fsf@gitster.dls.corp.google.com>
In-Reply-To
<CACRoPnRVjNseXPOxKe4_3rnKXmAYeA80iCHY3o8euWA=EUX6hA@mail.gmail.com>
Paul Tan <pyokagan@gmail.com> writes:
> I don't see how it feels iffy.
That is largely a taste thing.  And a good taste matters.

What is iffy is to use strbuf as an external interface between the implementation of the parse_opt_pass() API function and its users.

I would expect that no users of the parse_opt_pass() would ever take advantage of the fact that the value they are using is implemented as a strbuf, allowing them to use the family of functions in the strbuf API to further manipulate it. All users you added wants to use a plain vanilla string, and only because you used strbuf as the interface element, they have to say "var.buf" to get to the value they want to use.

> The purpose of using strbufs (and
> argv_arrays) is to avoid error-prone manual memory management.

It is a good idea to use strbuf as an implementation detail of parse_opt_pass() function. After all, safer, easier and efficient manipulation of strings is why we added the strbuf API to the system in the first place.

But it is a different matter to expose that as an API element when the recipient is not going to benefit.

In other words, callers of the API designed with a better taste would look more like this:

	static const char *opt_diffstat;
	static struct option pull_options[] = {
        	...
                { OPTION_CALLBACK, 'n', NULL, &opt_diffstat, NULL,
		  N_("do not show a diffstat at the end of the merge"),
		  PARSE_OPT_NOARG | PARSE_OPT_NONEG, parse_opt_pass
        	},
		...
	};
	static int run_merge(void)
        {
        	...
                if (opt_diffstat)
                	argv_array_push(&args, opt_diffstat);
		...
	}

That way, they do not have to define strbuf variables and dereference the value as opt_diffstat.buf.

And the implementation of the API element is not all that different from what you wrote:

	int parse_opt_pass(const struct options *opt, const char *arg, int unset)
	{
		struct strbuf sb = STRBUF_INIT;
		char **value = opt->value;
		if (*value)
                	free(*value);
		if (opt->long_name) {
                	strbuf_addf(&sb, "--%s%s",
                        	    unset ? "no-" : "",
                                    opt->long_name);
			...
                }
                ...
                *value = strbuf_detach(&sb, NULL);
		return 0;
	}
Previous: Paul TanNext: Paul Tan
Message 8 of 48 in “Make git-pull a builtin”
  1. 00/19 Make git-pull a builtinPaul Tan, Jun 3, 2015
  2. 01/19 parse-options-cb: implement parse_opt_pass_strbuf()Paul Tan, Jun 3, 2015
  3. Junio C HamanoJun 9, 2015
  4. 02/19 parse-options-cb: implement parse_opt_pass_argv_array()Paul Tan, Jun 3, 2015
  5. Stefan BellerJun 3, 2015
  6. Junio C HamanoJun 9, 2015
  7. Paul TanJun 10, 2015
  8. Junio C HamanoJun 10, 2015
  9. 03/19 argv-array: implement argv_array_pushv()Paul Tan, Jun 3, 2015
  10. 04/19 pull: implement skeletal builtin pullPaul Tan, Jun 3, 2015
  11. Junio C HamanoJun 10, 2015
  12. 05/19 pull: implement fetch + mergePaul Tan, Jun 3, 2015
  13. Junio C HamanoJun 9, 2015
  14. 06/19 pull: pass verbosity, --progress flags to fetch and mergePaul Tan, Jun 3, 2015
  15. Junio C HamanoJun 9, 2015
  16. 07/19 pull: pass git-merge's options to git-mergePaul Tan, Jun 3, 2015
  17. 08/19 pull: pass git-fetch's options to git-fetchPaul Tan, Jun 3, 2015
  18. Stefan BellerJun 3, 2015
  19. 09/19 pull: error on no merge candidatesPaul Tan, Jun 3, 2015
  20. Junio C HamanoJun 9, 2015
  21. Paul TanJun 13, 2015
  22. 10/19 pull: support pull.ff configPaul Tan, Jun 3, 2015
  23. Junio C HamanoJun 9, 2015
  24. 11/19 pull: check if in unresolved merge statePaul Tan, Jun 3, 2015
  25. Junio C HamanoJun 10, 2015
  26. Junio C HamanoJun 10, 2015
  27. Paul TanJun 10, 2015
  28. Junio C HamanoJun 10, 2015
  29. Paul TanJun 14, 2015
  30. 12/19 pull: fast-forward working tree if head is updatedPaul Tan, Jun 3, 2015
  31. 13/19 pull: implement pulling into an unborn branchPaul Tan, Jun 3, 2015
  32. Junio C HamanoJun 10, 2015
  33. 14/19 pull: set reflog messagePaul Tan, Jun 3, 2015
  34. 15/19 pull: teach git pull about --rebasePaul Tan, Jun 3, 2015
  35. Junio C HamanoJun 10, 2015
  36. Paul TanJun 10, 2015
  37. Junio C HamanoJun 10, 2015
  38. Paul TanJun 10, 2015
  39. Junio C HamanoJun 10, 2015
  40. Junio C HamanoJun 10, 2015
  41. 16/19 pull: configure --rebase via branch.<name>.rebase or pull.rebasePaul Tan, Jun 3, 2015
  42. 17/19 pull --rebase: exit early when the working directory is dirtyPaul Tan, Jun 3, 2015
  43. Kevin DaudtJun 3, 2015
  44. Kevin DaudtJun 10, 2015
  45. 18/19 pull --rebase: error on no merge candidate casesPaul Tan, Jun 3, 2015
  46. Stefan BellerJun 3, 2015
  47. 19/19 pull: remove redirection to git-pull.shPaul Tan, Jun 3, 2015
  48. Stefan BellerJun 3, 2015

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.