Re: [PATCH v5 05/10] add-interactive.c: implement status command

From: Slavica Djukic <slavicadj.ip2018@gmail.com>
To: Junio C Hamano <gitster@pobox.com>,
	Slavica Djukic via GitGitGadget <gitgitgadget@gmail.com>
Cc: git@vger.kernel.org, "Phillip Wood" <phillip.wood@dunelm.org.uk>,
	"Ævar Arnfjörð Bjarmason" <avarab@gmail.com>
Subject: Re: [PATCH v5 05/10] add-interactive.c: implement status command
Date: Fri, 1 Mar 2019 12:08:56 +0100	[thread overview]
Message-ID: <22f856ad-8248-7f1b-a720-636da6286e82@gmail.com> (raw)
In-Reply-To: <xmqq4l8vpj6x.fsf@gitster-ct.c.googlers.com>

On 22-Feb-19 11:11 PM, Junio C Hamano wrote:
> "Slavica Djukic via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
>> +static int parse_color_slot(const char *slot)
>> +{
>> +	if (!strcasecmp(slot, "prompt"))
>> +		return COLOR_PROMPT;
>> +	if (!strcasecmp(slot, "header"))
>> +		return COLOR_HEADER;
>> +	if (!strcasecmp(slot, "help"))
>> +		return COLOR_HELP;
>> +	if (!strcasecmp(slot, "error"))
>> +		return COLOR_ERROR;
> As these are "color.interactive.<name>", matching case-insensitively
> is the right thing to do.  Good.
>
> If we would end up increasing the number of slots, we may need to
> switch to LOOKUP_CONFIG(), but for just four of them, this will do.
>
>> @@ -313,3 +357,78 @@ static struct choices *list_and_choose(struct choices *data,
>>   		return NULL;
>>   	}
>>   }
>> +
>> +static struct choice *make_choice(const char *name )
> Excess SP before ')'?
>
>> +{
>> +	struct choice *choice;
> Style: have a blank line here to delineate decls (at the beginning
> of a function) and stmts (at all the decls).
>
>> +	FLEXPTR_ALLOC_STR(choice, name, name);
>> +	return choice;
>> +}
>> +
>> +static struct choice *add_choice(struct choices *choices, const char type,
>> +				 struct file_stat *file, struct command *command)
>> +{
>> +	struct choice *choice;
> Ditto here.
>
>> +	switch (type) {
>> +		case 'f':
> Style: in our codebase, switch and case are indented to the same
> level, with the body of each case arm indented one more level.
>
>> +			choice = make_choice(file->name);
>> +			choice->u.file.index.added = file->index.added;
>> +			choice->u.file.index.deleted = file->index.deleted;
>> +			choice->u.file.worktree.added = file->worktree.added;
>> +			choice->u.file.worktree.deleted = file->worktree.deleted;
> Would it make sense to make sure that all of file->index,
> u.file.index, file->wt, u.file.wt are exactly the same type of
> struct by introducing
>
> 	struct adddel { uintmax_t add, del; };
>
> in a very early part of the series, and embed that structure as a
> member in "struct choice" and "struct file_stat"?  That way, these
> assignments would become two structure assignments, that would be
> much easier to read.

Yes, definitely. I worked on this and it will be included in the next 
iteration.

>
>> +void add_i_status(void)
>> +{
>> +	struct collection_status *s;
>> +	struct list_and_choose_options opts = { 0 };
>> +	struct hashmap *map;
>> +	struct hashmap_iter iter;
>> +	struct choices choices = CHOICES_INIT;
>> +	struct file_stat *entry;
>> +	const char *modified_fmt = _("%12s %12s %s");
>> +	const char type = 'f';
>> +
>> +	opts.header = xmalloc(sizeof(char) * (HEADER_MAXLEN + 1));
>> +	snprintf(opts.header, HEADER_MAXLEN + 1, modified_fmt,
>> +		 _("staged"), _("unstaged"), _("path"));
> Is there aversion to use of strbuf among your mentors?
>
>
>> +	s = list_modified(the_repository, NULL);
>> +	if (s == NULL)
>> +		return;
>> +
>> +	map = &s->file_map;
>> +	hashmap_iter_init(map, &iter);
>> +	while ((entry = hashmap_iter_next(&iter))) {
>> +		add_choice(&choices, type, entry, NULL);
>> +	}
>> +
>> +	list_and_choose(&choices, &opts);
> In what order are these filenames given?  Whatever random order the
> hashmap happens to store them in?
>
> I vaguely recall in an earlier step the code used hashmap to collect
> but at the end produced a sorted list out of the final result.
> Shouldn't we be iterating over that sorted list instead?  Do we even
> need the hashmap at this point?

We actually don't need hashmap at this point, I've changed this so that
list_modified returns produced sorted list.
I've also applied all other suggestions in this message.
Thank you.

>> +	hashmap_free(&s->file_map, 1);
>> +	free(s);
>> +	free_choices(&choices);
> Did we just leak opt.header?
>
>> +}
>> diff --git a/add-interactive.h b/add-interactive.h
>> new file mode 100644
>> index 0000000000..8ef3d2e82b
>> --- /dev/null
>> +++ b/add-interactive.h
>> @@ -0,0 +1,8 @@
>> +#ifndef ADD_INTERACTIVE_H
>> +#define ADD_INTERACTIVE_H
>> +
>> +int add_i_config(const char *var, const char *value, void *cbdata);
>> +
>> +void add_i_status(void);
>> +
>> +#endif
>> diff --git a/builtin/add--helper.c b/builtin/add--helper.c
>> index 6a97f0e191..464d2245f3 100644
>> --- a/builtin/add--helper.c
>> +++ b/builtin/add--helper.c
>> @@ -1,6 +1,38 @@
>> +#include "add-interactive.h"
>>   #include "builtin.h"
>> +#include "config.h"
>> +#include "revision.h"
>> +
>> +static const char * const builtin_add_helper_usage[] = {
>> +	N_("git add-interactive--helper <command>"),
>> +	NULL
>> +};
>> +
>> +enum cmd_mode {
>> +	DEFAULT = 0,
>> +	STATUS
>> +};
>>   
>>   int cmd_add__helper(int argc, const char **argv, const char *prefix)
>>   {
>> +	enum cmd_mode mode = DEFAULT;
>> +
>> +	struct option options[] = {
>> +		OPT_CMDMODE(0, "status", &mode,
>> +			    N_("print status information with diffstat"), STATUS),
>> +		OPT_END()
>> +	};
>> +
>> +	git_config(add_i_config, NULL);
>> +	argc = parse_options(argc, argv, NULL, options,
>> +			     builtin_add_helper_usage,
>> +			     PARSE_OPT_KEEP_ARGV0);
>> +
>> +	if (mode == STATUS)
>> +		add_i_status();
>> +	else
>> +		usage_with_options(builtin_add_helper_usage,
>> +				   options);
>> +
>>   	return 0;
>>   }