From: Glen Choo <chooglen@google.com>
To: Phillip Wood <phillip.wood123@gmail.com>,
phillip.wood@dunelm.org.uk,
Glen Choo via GitGitGadget <gitgitgadget@gmail.com>,
git@vger.kernel.org
Cc: "Jonathan Tan" <jonathantanmy@google.com>,
"Ævar Arnfjörð Bjarmason" <avarab@gmail.com>,
"Emily Shaffer" <nasamuffin@google.com>
Subject: Re: [PATCH v2 03/14] (RFC-only) config: add kvi arg to config_fn_t
Date: Fri, 09 Jun 2023 16:19:02 -0700 [thread overview]
Message-ID: <kl6l3530gr6h.fsf@chooglen-macbookpro.roam.corp.google.com> (raw)
In-Reply-To: <1ca04355-8faf-e7f9-051f-a6a568ff99c5@gmail.com>
Phillip Wood <phillip.wood123@gmail.com> writes:
> On 02/06/2023 17:46, Glen Choo wrote:
>> Phillip Wood <phillip.wood123@gmail.com> writes:
>>> As an aside, I think we'd also want a couple of helpers for matching
>>> keys so we can just write kvi_match_key(kvi, "user.name") or
>>> kvi_skip_key_prefix(kvi, "core.", &p) rather than having to extract the
>>> key name first.
>>
>> Yes, and that would also abstract over implementation details like
>> matching keys using strcasecmp() and not strcmp(). For reasons like
>> this, I think your proposal paves the way for a harder-to-misuse API.
>>
>> I still have some nagging, probably irrational fear that consolidating
>> all of the config_fn_t args is trickier to manage than adding a single
>> key_value_info arg. It definitely *sounds* trickier, but I can't really
>> think of a real downside.
>>
>> Maybe I just have to try it and send the result for others to consider.
>
> That's probably the best way to see if it is an improvement - you can
> always blame me if it turns out not to be! Hopefully it isn't too much
> work to add enough api to be able to convert a couple of config_fn_t
> functions to see how it pans out.
>
I was preparing a reroll that incorporated some of the suggestions here,
but when I was done, I didn't really feel like it was better (or at the
very least, worth the extra churn) compared to the original approach.
The 3 versions I was comparing are:
1. Adding a new "struct config_context" arg to config_fn_t that only
contains a .kvi member. This is basically the approach in v1 and v2
(adding an extra args) but wrapping it in a "struct config_context" so
that we can grow it later.
2. Stuffing "key" and "value" into "struct key_value_info" and removing
"key" and "value" from config_fn_t.
3. Creating a "struct config_context" that contains "key", "value" and
"struct key_value_info" as separate members, and removing
"key" and "value" from config_fn_t.
I didn't try using opaque getters - I think that's nice to have, but
isn't crucial to this series and introduces a lot of churn.
I've uploaded an implementation of option 3. [1]. The result is quite
ugly in some places. In particular, config callbacks are used to calling
other config callbacks with slightly different args (e.g. massaging the
"key" but keeping the "value" the same, see patches 3,5/14 for
examples.), so to make this work, I ended up creating a new
"config_context", copying over the relevant members, then changing the
members that need to be different. I didn't pursue the refactor to its
end state (e.g. some of the CLI config machinery in the earlier patches
and git_config_int()/ functions that end in die_bad_number() in later
patches, take "key", "value", "key_value_info" as separate args instead
of a single "config_context" arg) so I haven't reaped all of the
benefits, but I thought the range-diff was getting churny enough that I
wasn't interested in pursuing the idea further.
I toyed with option 2 for a bit. Unsurprisingly, it requires even more
copying of members, though in some ways, it's better than option 3
because callers only have to initialize one "struct key_value_info"
instead of "struct key_value_info" _and_ "struct config_context".
However, it requires reworking the machinery quite a lot because
"key_value_info" is currently used _just_ for config source information,
and there are several cases where we need to propagate the config source
information separately from the key/value (e.g. the configsets store
"key" in a hashmap key, "value" in a string_list_item.string, and
"key_value_info" in a string_list_item.util). I'm convinced that this is
better interface-wise than option 3, but the required effort is high
enough that I don't think we should do this unless we see a real need
for it.
With that in mind, I'm tempted to continue with option 1 for now, but
let me know if I'm missing something.
[1] https://github.com/git/git/compare/master...chooglen:git:config/no-global-combined-kv-context?expand=1
next prev parent reply other threads:[~2023-06-09 23:19 UTC|newest]
Thread overview: 115+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-21 19:13 [PATCH 00/14] [RFC] config: remove global state from config iteration Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 01/14] config.c: introduce kvi_fn(), use it for configsets Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 02/14] config.c: use kvi for CLI config Glen Choo via GitGitGadget
2023-05-01 11:06 ` Ævar Arnfjörð Bjarmason
2023-04-21 19:13 ` [PATCH 03/14] config: use kvi for config files Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 04/14] config: add kvi.path, use it to evaluate includes Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 05/14] config: pass source to config_parser_event_fn_t Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 06/14] config: inline git_color_default_config Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 07/14] urlmatch.h: use config_fn_t type Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 08/14] (RFC-only) config: add kvi arg to config_fn_t Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 09/14] (RFC-only) config: apply cocci to config_fn_t implementations Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 10/14] (RFC-only) config: finish config_fn_t refactor Glen Choo via GitGitGadget
2023-05-01 11:19 ` Ævar Arnfjörð Bjarmason
2023-05-05 21:07 ` Jonathan Tan
2023-05-09 22:46 ` Glen Choo
2023-05-11 16:21 ` Jonathan Tan
2023-05-08 21:00 ` Glen Choo
2023-04-21 19:13 ` [PATCH 11/14] config: remove current_config_(line|name|origin_type) Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 12/14] config: remove current_config_scope() Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 13/14] config: pass kvi to die_bad_number() Glen Choo via GitGitGadget
2023-04-21 19:13 ` [PATCH 14/14] config: remove config_reader from configset_add_value Glen Choo via GitGitGadget
2023-05-30 18:41 ` [PATCH v2 00/14] [RFC] config: remove global state from config iteration Glen Choo via GitGitGadget
2023-05-30 18:41 ` [PATCH v2 01/14] config: inline git_color_default_config Glen Choo via GitGitGadget
2023-05-30 18:42 ` [PATCH v2 02/14] urlmatch.h: use config_fn_t type Glen Choo via GitGitGadget
2023-05-30 18:42 ` [PATCH v2 03/14] (RFC-only) config: add kvi arg to config_fn_t Glen Choo via GitGitGadget
2023-06-01 9:50 ` Phillip Wood
2023-06-01 16:22 ` Glen Choo
2023-06-02 9:54 ` Phillip Wood
2023-06-02 16:46 ` Glen Choo
2023-06-05 9:38 ` Phillip Wood
2023-06-09 23:19 ` Glen Choo [this message]
2023-05-30 18:42 ` [PATCH v2 04/14] (RFC-only) config: apply cocci to config_fn_t implementations Glen Choo via GitGitGadget
2023-05-30 18:42 ` [PATCH v2 05/14] (RFC-only) config: finish config_fn_t refactor Glen Choo via GitGitGadget
2023-06-01 22:17 ` Jonathan Tan
2023-05-30 18:42 ` [PATCH v2 06/14] config.c: pass kvi in configsets Glen Choo via GitGitGadget
2023-06-01 22:21 ` Jonathan Tan
2023-05-30 18:42 ` [PATCH v2 07/14] config: provide kvi with config files Glen Choo via GitGitGadget
2023-06-01 22:41 ` Jonathan Tan
2023-06-01 23:54 ` Jonathan Tan
2023-05-30 18:42 ` [PATCH v2 08/14] builtin/config.c: test misuse of format_config() Glen Choo via GitGitGadget
2023-05-30 18:42 ` [PATCH v2 09/14] config.c: provide kvi with CLI config Glen Choo via GitGitGadget
2023-06-01 23:35 ` Jonathan Tan
2023-06-02 17:26 ` Glen Choo
2023-05-30 18:42 ` [PATCH v2 10/14] trace2: plumb config kvi Glen Choo via GitGitGadget
2023-06-01 23:38 ` Jonathan Tan
2023-05-30 18:42 ` [PATCH v2 11/14] config: pass kvi to die_bad_number() Glen Choo via GitGitGadget
2023-06-01 23:48 ` Jonathan Tan
2023-06-02 17:23 ` Glen Choo
2023-05-30 18:42 ` [PATCH v2 12/14] config.c: remove config_reader from configsets Glen Choo via GitGitGadget
2023-05-30 18:42 ` [PATCH v2 13/14] config: add kvi.path, use it to evaluate includes Glen Choo via GitGitGadget
2023-06-02 0:06 ` Jonathan Tan
2023-05-30 18:42 ` [PATCH v2 14/14] config: pass source to config_parser_event_fn_t Glen Choo via GitGitGadget
2023-06-02 0:08 ` Jonathan Tan
2023-06-02 17:20 ` Glen Choo
2023-06-20 19:43 ` [PATCH v3 00/12] config: remove global state from config iteration Glen Choo via GitGitGadget
2023-06-20 19:43 ` [PATCH v3 01/12] config: inline git_color_default_config Glen Choo via GitGitGadget
2023-06-20 21:01 ` Junio C Hamano
2023-06-20 19:43 ` [PATCH v3 02/12] urlmatch.h: use config_fn_t type Glen Choo via GitGitGadget
2023-06-20 21:02 ` Junio C Hamano
2023-06-20 19:43 ` [PATCH v3 03/12] config: add ctx arg to config_fn_t Glen Choo via GitGitGadget
2023-06-20 19:43 ` [PATCH v3 04/12] config.c: pass ctx in configsets Glen Choo via GitGitGadget
2023-06-20 21:19 ` Junio C Hamano
2023-06-20 19:43 ` [PATCH v3 05/12] config: pass ctx with config files Glen Choo via GitGitGadget
2023-06-20 19:43 ` [PATCH v3 06/12] builtin/config.c: test misuse of format_config() Glen Choo via GitGitGadget
2023-06-20 21:35 ` Junio C Hamano
2023-06-20 23:06 ` Glen Choo
2023-06-23 20:32 ` Jonathan Tan
2023-06-24 1:31 ` Jeff King
2023-06-28 17:28 ` Glen Choo
2023-06-20 19:43 ` [PATCH v3 07/12] config.c: pass ctx with CLI config Glen Choo via GitGitGadget
2023-06-23 20:35 ` Jonathan Tan
2023-06-23 21:41 ` Glen Choo
2023-06-20 19:43 ` [PATCH v3 08/12] trace2: plumb config kvi Glen Choo via GitGitGadget
2023-06-23 20:40 ` Jonathan Tan
2023-06-20 19:43 ` [PATCH v3 09/12] config: pass kvi to die_bad_number() Glen Choo via GitGitGadget
2023-06-20 19:43 ` [PATCH v3 10/12] config.c: remove config_reader from configsets Glen Choo via GitGitGadget
2023-06-23 20:57 ` Jonathan Tan
2023-06-23 21:33 ` Junio C Hamano
2023-06-20 19:43 ` [PATCH v3 11/12] config: add kvi.path, use it to evaluate includes Glen Choo via GitGitGadget
2023-06-20 19:43 ` [PATCH v3 12/12] config: pass source to config_parser_event_fn_t Glen Choo via GitGitGadget
2023-06-20 21:46 ` Junio C Hamano
2023-06-21 21:46 ` [PATCH v3 00/12] config: remove global state from config iteration Junio C Hamano
2023-06-21 23:06 ` Glen Choo
2023-06-23 21:02 ` Jonathan Tan
2023-06-23 21:33 ` Junio C Hamano
2023-06-23 21:45 ` Glen Choo
2023-06-26 18:11 ` [PATCH v4 " Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 01/12] config: inline git_color_default_config Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 02/12] urlmatch.h: use config_fn_t type Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 03/12] config: add ctx arg to config_fn_t Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 04/12] config.c: pass ctx in configsets Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 05/12] config: pass ctx with config files Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 06/12] builtin/config.c: test misuse of format_config() Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 07/12] config.c: pass ctx with CLI config Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 08/12] trace2: plumb config kvi Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 09/12] config: pass kvi to die_bad_number() Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 10/12] config.c: remove config_reader from configsets Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 11/12] config: add kvi.path, use it to evaluate includes Glen Choo via GitGitGadget
2023-06-26 18:11 ` [PATCH v4 12/12] config: pass source to config_parser_event_fn_t Glen Choo via GitGitGadget
2023-06-26 20:45 ` [PATCH v4 00/12] config: remove global state from config iteration Junio C Hamano
2023-06-28 19:26 ` [PATCH v5 00/11] " Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 01/11] config: inline git_color_default_config Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 02/11] urlmatch.h: use config_fn_t type Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 03/11] config: add ctx arg to config_fn_t Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 04/11] config.c: pass ctx in configsets Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 05/11] config: pass ctx with config files Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 06/11] config.c: pass ctx with CLI config Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 07/11] trace2: plumb config kvi Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 08/11] config: pass kvi to die_bad_number() Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 09/11] config.c: remove config_reader from configsets Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 10/11] config: add kvi.path, use it to evaluate includes Glen Choo via GitGitGadget
2023-06-28 19:26 ` [PATCH v5 11/11] config: pass source to config_parser_event_fn_t Glen Choo via GitGitGadget
2023-06-28 22:23 ` [PATCH v5 00/11] config: remove global state from config iteration Jonathan Tan
2023-06-28 22:47 ` Junio C Hamano
2023-07-11 18:41 ` Phillip Wood
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=kl6l3530gr6h.fsf@chooglen-macbookpro.roam.corp.google.com \
--to=chooglen@google.com \
--cc=avarab@gmail.com \
--cc=git@vger.kernel.org \
--cc=gitgitgadget@gmail.com \
--cc=jonathantanmy@google.com \
--cc=nasamuffin@google.com \
--cc=phillip.wood123@gmail.com \
--cc=phillip.wood@dunelm.org.uk \
/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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).