Locating identifiers in tactics (including ltac2, ssr) #13103

herbelin · 2020-09-27T15:05:34Z

Kind: "cleanup"

This is part of the project started in #11842.

We use located identifiers in tactics in anticipation of identifying Prim.identref and Prim.ident, with the latter eventually taking a location by default and the former being eventually deprecated.

Tactics intro_move, rename_hyp, instantiate_by_name, fix/cofix, destruct, ... now take a located identifier. This is a compromise in the sense that internal calls to these tactics need to add a dummy loc just to comply to the extended API. On the other side, having the location at hand in error messages is desirable when the call directly comes from the user (for instance, intros H may now hightlight H, though there are still some issues, as indicated in #13100, which make that intro H does not highlight errors on H).

The first commit is shared with #13102 (it does not matter which PR comes first).

The four next commits are about tactics but do not concern interpretation, thus, in some sense, they could have been alternatively in #13100 (for HypLocation) or #13102 (for Derive Inversion or Ltac definitions).

The last six commits (7th to 12th) rely on the support for located identifiers in ARGUMENT EXTEND and TACTIC EXTEND, which is provided by the 6th commit.

In ssrparser, some harmonization could be done, e.g. using Id.t CAst.t in some places rather than Id.t Loc.located (e.g. in SsrHyp). I can do it if there is a will to, or the work can also be shared, if this is what is preferred.

In Ltac2, I did not change the parsing rules so that they possibly pass located identifiers to tactics. I may try to do it, if there is a demand.

coqbot-app · 2020-09-27T16:21:58Z

The job library:ci-fiat_crypto_legacy has failed in allow failure mode
ping @JasonGross

coqbot-app · 2020-10-04T14:08:33Z

The job library:ci-fiat_crypto_legacy has failed in allow failure mode
ping @JasonGross

herbelin · 2020-10-12T07:34:05Z

The PR removed induction 0 and destruct 0 and this is the cause of failures (or at least the main cause). Some of these uses are actually legitimate. I'm proposing to implement induction and destruct without an argument to mean what the "hackish" induction 0 and destruct 0 did.

gares · 2020-10-12T09:40:23Z

what the "hackish" induction 0 and destruct 0 did.

What did they do?

ppedrot · 2020-10-13T07:16:15Z

FTR, I really don't like this PR. It is adding a lot of debugging information in the wrong place, namely the toplevel value of something that is not an AST, thus polluting the runtime with a lot of irrelevant data. Most of the time one does not even want to know about this debug data, especially when it comes from another file, e.g. when declaring toplevel tactics in some general purpose library. Maybe for Ltac1 it's a necessity because there is no clear separation between idents and other similarly shaped data, but I firmly believe that for Ltac2 this should be handled at the level of functions via backtraces, and definitely not at the level of data for such generic purpose types (user-facing ASTs are another matter, though). As such, except if there is a very good argument in favour of this PR I'd prefer a statu quo.

gares · 2020-10-13T07:55:32Z

plugins/ssr/ssrparser.mlg

-    let s, id' = interp_wit wit_ident ist gl id in
-    s, Id (SsrHyp (loc, id'))
+    let s, id' = interp_wit wit_identref ist gl (CAst.make ?loc id) in
+    s, Id (SsrHyp (loc, id'.CAst.v))


I guess this diff hunk supports @ppedrot thesis. Here the loc is already in the AST, the copy in the ident is just getting in the way. It would feel more natural to pass loc to interp_wit frankly.

I would say that it is expected that interp_wit takes a loc since it goes from user-facing syntax to "abstract" syntax (the first line of the diff). What is more questionable (and what @ppedrot questions, afaiu) is that interp_wit continues to return the loc (the second line of the diff).

What I meant is that if you don't put the loc in the ident in the fist place, then interp_wit is not going to return it, you can still pass the loc as you pass ist and gl if you need to

herbelin · 2020-10-13T11:02:53Z

I don't like either to pollute a library part (tactics.ml) with user input data, especially since the library can indeed be called from other files w/o any relation to a user input (this is actually why I avoided to push such locations to tactics.ml in the past). On the other side, it is legitimate to be able to locate errors, so the question of how to do it is open. What would it mean to use "backtraces"? That the loc is attached afterwards by the caller?

For instance, would something like following be acceptable for you: the tactic in tactics.ml raises an exception say (AddLocOfIdent (id,exn)) (or attach id to the info of the exn), and the caller reraises exnwith the loc ofid`?

In any case, I'm ok with the statu quo (the error messages are already designed on the assumption that they do not know the loc and thus repeat the name of the identifier), but I'd really be interested that the underlying issue is not ignored.

The PR has however more than just changing tactics.ml. There is a part which locates the identifiers as AST and this should be kept, up e.g. to have:

val wit_identref : (lident, lident, Id.t) genarg_type

instead of

val wit_identref : lident uniform_genarg_type

Is this OK?

ppedrot · 2020-10-13T12:07:37Z

Is this OK?

Nah, that's also runtime polluting.

That the loc is attached afterwards by the caller?

More precisely, only attaching such data when inside a script. This could be done by inserting locating wrappers inside the AST in non-strict mode, for instance.

herbelin · 2020-10-13T12:28:48Z

Nah, that's also runtime polluting.

Do you mean that you have a solution to avoid loc in pretyping.ml and friends?

This could be done by inserting locating wrappers inside the AST in non-strict mode, for instance.

My imagination is unfortunately not big enough to visualize what you have in mind...

In passing, we use the location also for intros, etc.

…arguments.

coqbot-app · 2020-10-28T16:20:34Z

The job library:ci-fiat_crypto_legacy has failed in allow failure mode
ping @JasonGross

herbelin · 2020-10-28T19:18:56Z

This could be done by inserting locating wrappers inside the AST in non-strict mode, for instance.

If I call e.g. intros x [|y], how this wrapper would know that it is x or y or [|y] or the whole intros x [|y] which has to be highlighted? The information is only known from the code of intros and that would be a complicated protocol to inform the caller of which exact subterm of the arguments is to be highlighted.

gares · 2020-11-16T13:13:07Z

My understanding is that this is not ready and some discussion (esp with @ppedrot) has to happen to devise the design

coqbot-app · 2021-07-28T14:03:37Z

The "needs: rebase" label was set more than 30 days ago. If the PR is not rebased in 30 days, it will be automatically closed.

coqbot-app · 2021-09-13T02:13:28Z

This PR was not rebased after 30 days despite the warning, it is now closed.

herbelin added kind: cleanup Code removal, deprecation, refactorings, etc. part: tactics part: parser The parser (also called gramlib, forked from camlp5) labels Sep 27, 2020

herbelin added this to the 8.13+beta1 milestone Sep 27, 2020

herbelin requested review from a team as code owners September 27, 2020 15:05

herbelin force-pushed the master+located-identifiers-tactics branch from ba24eb9 to d26c3f7 Compare September 27, 2020 15:38

herbelin requested a review from a team as a code owner September 27, 2020 15:38

herbelin mentioned this pull request Sep 29, 2020

Towards better grammar entry names: renaming "identref" into "lident"? #11842

Closed

github-actions bot added the needs: rebase Should be rebased on the latest master to solve conflicts or have a newer CI run. label Sep 30, 2020

herbelin force-pushed the master+located-identifiers-tactics branch from d26c3f7 to 2361efa Compare October 4, 2020 13:25

coqbot-app bot removed the needs: rebase Should be rebased on the latest master to solve conflicts or have a newer CI run. label Oct 4, 2020

gares reviewed Oct 13, 2020

View reviewed changes

github-actions bot added the needs: rebase Should be rebased on the latest master to solve conflicts or have a newer CI run. label Oct 15, 2020

herbelin force-pushed the master+located-identifiers-tactics branch from 2361efa to 41a1c32 Compare October 15, 2020 23:09

herbelin added 11 commits October 28, 2020 16:36

Using located identifiers for HypLocation.

7777367

Using located identifiers when parsing inversion lemmas commands.

220080a

Using preident or located identifiers when parsing Ltac.

8759d44

Adding Ltac interning/interpreting/subst/printing for identref.

f521337

Using located identifiers in tactic rename_hyp.

f8b002f

Using a located identifier in tactic instantiate_tac_by_name.

bcb3147

Using located identifiers in tactic intro_move.

be1a48b

In passing, we use the location also for intros, etc.

Using a located identifier in tactics fix and cofix.

e0af06f

Using located identifiers in NamedHyp/AnonHyp tactic arguments.

cd5acea

Using located identifiers when parsing a couple of tactics or tactic …

96b55bd

…arguments.

Using located identifiers in ssrvernac and ssrparser grammars.

5facd93

herbelin force-pushed the master+located-identifiers-tactics branch from 41a1c32 to 5facd93 Compare October 28, 2020 15:46

coqbot-app bot removed the needs: rebase Should be rebased on the latest master to solve conflicts or have a newer CI run. label Oct 28, 2020

github-actions bot added the needs: rebase Should be rebased on the latest master to solve conflicts or have a newer CI run. label Nov 5, 2020

herbelin mentioned this pull request Nov 5, 2020

Locating identifiers in a few plugin commands #13102

Merged

gares added the needs: discussion Further discussion is needed. label Nov 16, 2020

gares modified the milestones: 8.13+beta1, 8.14+beta1 Nov 16, 2020

ppedrot modified the milestones: 8.14+rc1, 8.15+rc1 Jun 7, 2021

coqbot-app bot added the stale This PR will be closed unless it is rebased. label Jul 28, 2021

coqbot-app bot closed this Sep 13, 2021

coqbot-app bot removed this from the 8.15+rc1 milestone Sep 13, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Locating identifiers in tactics (including ltac2, ssr) #13103

Locating identifiers in tactics (including ltac2, ssr) #13103

herbelin commented Sep 27, 2020 •

edited

Loading

coqbot-app bot commented Sep 27, 2020

coqbot-app bot commented Oct 4, 2020

herbelin commented Oct 12, 2020

gares commented Oct 12, 2020

ppedrot commented Oct 13, 2020

gares Oct 13, 2020

herbelin Oct 13, 2020

gares Oct 13, 2020

herbelin commented Oct 13, 2020

ppedrot commented Oct 13, 2020

herbelin commented Oct 13, 2020

coqbot-app bot commented Oct 28, 2020

herbelin commented Oct 28, 2020

gares commented Nov 16, 2020

coqbot-app bot commented Jul 28, 2021

coqbot-app bot commented Sep 13, 2021

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Locating identifiers in tactics (including ltac2, ssr) #13103

Locating identifiers in tactics (including ltac2, ssr) #13103

Conversation

herbelin commented Sep 27, 2020 • edited Loading

coqbot-app bot commented Sep 27, 2020

coqbot-app bot commented Oct 4, 2020

herbelin commented Oct 12, 2020

gares commented Oct 12, 2020

ppedrot commented Oct 13, 2020

gares Oct 13, 2020

Choose a reason for hiding this comment

herbelin Oct 13, 2020

Choose a reason for hiding this comment

gares Oct 13, 2020

Choose a reason for hiding this comment

herbelin commented Oct 13, 2020

ppedrot commented Oct 13, 2020

herbelin commented Oct 13, 2020

coqbot-app bot commented Oct 28, 2020

herbelin commented Oct 28, 2020

gares commented Nov 16, 2020

coqbot-app bot commented Jul 28, 2021

coqbot-app bot commented Sep 13, 2021

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

herbelin commented Sep 27, 2020 •

edited

Loading