Re: [PATCH 5/5] u-string-list: move "remove duplicates" test to "u-string-list.c"

[Patrick Steinhardt <ps@pks.im>]

On Tue, Apr 22, 2025 at 10:55:23PM +0800, shejialuo wrote:
> diff --git a/t/helper/test-string-list.c b/t/helper/test-string-list.c
> index 262b28c599..6be0cdb8e2 100644
> --- a/t/helper/test-string-list.c
> +++ b/t/helper/test-string-list.c
> @@ -1,48 +1,9 @@
> -#define DISABLE_SIGN_COMPARE_WARNINGS
> -
>  #include "test-tool.h"
>  #include "strbuf.h"
>  #include "string-list.h"
>  
> -/*
> - * Parse an argument into a string list.  arg should either be a
> - * ':'-separated list of strings, or "-" to indicate an empty string
> - * list (as opposed to "", which indicates a string list containing a
> - * single empty string).  list->strdup_strings must be set.
> - */
> -static void parse_string_list(struct string_list *list, const char *arg)
> -{
> -	if (!strcmp(arg, "-"))
> -		return;
> -
> -	(void)string_list_split(list, arg, ':', -1);
> -}
> -
> -static void write_list_compact(const struct string_list *list)
> -{
> -	int i;
> -	if (!list->nr)
> -		printf("-\n");
> -	else {
> -		printf("%s", list->items[0].string);
> -		for (i = 1; i < list->nr; i++)
> -			printf(":%s", list->items[i].string);
> -		printf("\n");
> -	}
> -}
> -
>  int cmd__string_list(int argc, const char **argv)
>  {
> -	if (argc == 3 && !strcmp(argv[1], "remove_duplicates")) {
> -		struct string_list list = STRING_LIST_INIT_DUP;
> -
> -		parse_string_list(&list, argv[2]);
> -		string_list_remove_duplicates(&list, 0);
> -		write_list_compact(&list);
> -		string_list_clear(&list, 0);
> -		return 0;
> -	}
> -
>  	if (argc == 2 && !strcmp(argv[1], "sort")) {
>  		struct string_list list = STRING_LIST_INIT_NODUP;
>  		struct strbuf sb = STRBUF_INIT;

I'm a bit surprised that the patch series stops after this patch given
that the only remaining subcommand is "sort". Is there a specific reason
why you don't also convert that function? If you did we could declare
victory by deleting the whole "test-helper string-list" subcommand.

It seems that the only reason is p0071, where we benchmark performance
of sorting. I dunno... that one is of course not a good fit for our unit
testing framework. But it's a bit sad that we cannot remove the whole
infra only because of a performance test that nobody ever runs in the
first place.

Patrick