Just some tbl things. by Drup · Pull Request #1699 · ocaml/ocaml

Drup · 2018-04-05T19:06:31Z

I always felt that the Tbl module was a ~~gross~~slight violation of the DRY principle, so this patchset removes it.
On the way, I removed almost all the duplicated (String|Int)(Map|Set) in the compiler, and a few others.

This is my small contribution towards making the compiler feel like one code base, instead of the collision of five different ones. ;)

Details on `Tbl`

Tbl is basically a polymorphic Map module, using the polymorphic comparison with a few more hacks (such as Tbl.find_str which hardcodes the string comparison for speed. It is used for some string tables ... but not all of them). It has the exact same structure and semantics as Map, making it completely redundant.

On top of that, relying on the generic Tbl make local modifications fairly painful and invasive. For example, in the eliom patches, names are a couple of a string and a "location" (in eliom vocabulary) with a custom comparison function. With this patch, it can be implemented as a simple change in the implementation of Env.NameMap.

Patchset

The first 10 commits should be fairly consensual. They were mostly done through paste and replace.

The last 3 commits, particularly 484c79b , might be less consensual: Tbl made its way into the format for cmo files though the symbol tables for global variables and primitives. I replaced them by an Ident.Map (for global ids) and a StringMap (for primitives). I didn't bump the magic number (yet).

I'm not sure if I need a change entry for this.

Perfs

I don't expect any perf differences (the "hot" paths were already using Tbl.find_str). My non-scientific benchmark show a very minor speedup (around 1%, so probably noise). Is there some command to measure performance changes of the compiler on itself?

Drup · 2018-04-05T19:38:52Z

I do not understand travis' failure. It builds/tests fine locally and the various depend are up-to-date.

Octachron · 2018-04-05T19:47:04Z

There is a .depend file hidden in ocamldoc/stdlib_non_prefixed, it seems that this the one responsible for the failure.

mshinwell · 2018-04-06T08:27:09Z

I haven't got time to read this in detail right now, but will aim to do so in due course. I agree that cleanup here would be beneficial. This work seems related to some changes I am working on to integrate the utils/identifiable.ml* functionality directly into the standard library. I will write about that within the next few weeks.

nojb · 2018-04-06T08:31:15Z

I didn't bump the magic number (yet).

This is not necessary. Magic numbers will be bumped unconditionally for each release.

dra27 · 2018-04-06T08:38:27Z

At a glance, this looks good to me (although, if I may, I think the word gross should possibly only be used when describing one's own work!).

It's a basically internal change, so I agree it doesn't require a changes entry, but that does mean the work only gets credited in Git history, so feel free to add one if you'd like!

Drup · 2018-04-06T09:22:53Z

This work seems related to some changes I am working on to integrate the utils/identifiable.ml* functionality directly into the standard library.

Indeed! Making relevant data-structures in the compiler into Identifiable.S would also be very nice. I almost started doing that when I realized that Identifiable also include printing, and printing for the source/type language is not so trivial (the printers are defined somewhere else, and require lot's of machinery).

It's a basically internal change, so I agree it doesn't require a changes entry, but that does mean the work only gets credited in Git history, so feel free to add one if you'd like!

Ok, I will add one then.

mshinwell

I've made some comments and suggestions. This is good work.

mshinwell · 2018-04-09T14:05:09Z

 (* Introduction of closures, uncurrying, recognition of direct calls *)

 open Misc
+open Numbers


Instead of a whole open, I would just use: module Int = Numbers.Int here.

mshinwell · 2018-04-09T14:05:30Z

 open Switch
 open Clambda

+module VarMap = Ident.Map


I think this is confusing ("ident"s versus "var"s). How about just replacing VarMap with Ident.Map throughout? That seems fine to me, and more clear in fact.

mshinwell · 2018-04-09T14:06:07Z

-    type t = int
-    let compare (x:t) y = compare x y
-  end)
+module IntSet = Numbers.Int.Set


Here, I would use module Int = Numbers.Int again, then this alias can be removed. Just use Int.Set instead of IntSet.

mshinwell · 2018-04-09T14:06:22Z

   sequentialization. *)

 open Misc
+open Numbers


Same comment as above.

mshinwell · 2018-04-09T14:07:43Z

 open Asttypes
 open Lambda

+module VarMap = Ident.Map


Same comment as above.

mshinwell · 2018-04-09T14:19:23Z

 end

-module PathMap = Map.Make(Path)
+module NameMap = Misc.StringMap


Names of what?

Name, as in the individual parts in an Longindent.t. I would like to call them ident, but that clashes with the individual parts of a Path.t. I insist on having a local redefinition on that one, as it's the original motivation for this whole patch. But I agree, the name is bad.

mshinwell · 2018-04-09T14:21:18Z

  | Field_class _ -> "class"
  | Field_classtype _ -> "class type"

+(** Map indexed field names. Allow to preserve namespaces. *)


This comment isn't very clear---can you reword?

The new one is :

(** Map indexed by both field types and names. This avoids name clashes between different sorts of fields such as values and types. *)

mshinwell · 2018-04-09T14:24:03Z

-    | Papply _ -> false
+    | Path.Pident _ -> true
+    | Path.Pdot(p, _, _) -> no_apply p
+    | Path.Papply _ -> false


FWIW, using disambiguation helps a lot for this kind of thing:

let can_alias env path = let rec no_apply (path : Path.t) = match path with | Pident _ -> true | Pdot(p, _, _) -> no_apply p | Papply _ -> false in ...

Unfortunately, this part of the compiler has warning 40 turned on (and fatal). I think changing default warnings in the implementation of the typechecker shouldn't be discussed in this PR.

mshinwell · 2018-04-09T14:24:49Z

+  let compare = compare
+end
+module Set = Set.Make(T)
+module Map = Map.Make(T)


You could probably use Identifiable.Make for this instead.

Identifiable.Make require (two) printers and an equal function. I would prefer to leave that for later.

mshinwell · 2018-04-09T14:26:41Z

 exception Error_forward of Location.error

+(** Map indexed by type variable names. *)
+module VarMap = Misc.StringMap


Here is an example of why I recommended not having the VarMap alias up above in closure.ml: here the same alias means something completely different! (And it's not like it's a one-letter alias.)

I removed all instances of VarMap outside of the typechecker and renamed that one as TyVarMap, which should make things clearer.

mshinwell · 2018-04-09T14:32:33Z

Oh, and please add a Changes entry, in the "internal" section. This is a fairly large diff even if mostly straightforward.

Drup · 2018-04-10T14:08:54Z

@mshinwell Thanks for the detailed review! I think I addressed most of your comments. I also fixed the change entry and rebased.

There are a few things (cleanup the 3 extended stdlibs, add even more identifiable everywhere) that would be very beneficial, but I would prefer to do them later, as the risk of conflict between this patchset and other changes is quite high.

Drup · 2018-04-10T14:22:07Z

While replacing ocamltest's StringMap and Sets, I realized the function stringset_of_string was unused. So I removed it and simply use aliases to Map.Make's output.

Drup · 2018-06-28T23:38:53Z

I rebased and added some fixed related to various recent additions.

I think I already addressed all of @mshinwell 's comment. Is there a remaining blocker?

I plan to squash things before the merge.

damiendoligez · 2018-06-29T15:21:10Z

You don't have to squash, just state that it should be squashed and we just use GitHub's "Squash and merge" option.

gasche · 2018-06-29T16:12:01Z

"Squash and merge" is generally terrible except for PRs that obviously fit in just one commit. If it's a series of nicely separated commits ("Remove Tbl in Foo" fits the pattern) plus some junk, you have to squash manually.

On the other hand, @Drup is making a mistake by not squashing right away. He is trying to save his time, but introduces complexity, delay, communication needs and mistake points in the process instead. One might even go as far as saying as he deserves to get his entry squashed!

Drup · 2018-06-29T16:19:07Z

@gasche I'm going to quote @mshinwell In another PR of mine:

Please ensure this isn't squashed (or similar) before merging as it will be difficult to review.

New commits are not squashed immediately with the old ones so that reviewers don't have to make high order diffs between stack of patchs. Unfortunately, we're not using Iron for OCaml PR reviews.

gasche · 2018-06-29T16:23:18Z

I would disagree with Mark on this one, but given that he is the one who should give you his seal of approval, it makes sense to please him.

xavierleroy · 2018-06-29T16:26:27Z

"Squash and merge" is generally the right thing to do, because most PRs have an uninteresting history and uninteresting commit message, e.g. "Typos", "Going home", or "Add a Changes entry". Indeed this PR has one commit titled "Really?!" and another titled "Depend". I don't want to see this nor the 18 intermediate states of this PR in OCaml's commit log. I want to see one entry in the log with one message that summarizes what the PR does in its final state.

dra27 · 2018-06-30T16:59:08Z

If you want to tidy the history, then I'm sure @mshinwell can be persuaded to merge a clean commit history! However, before this is merged, please could check-typo violations be dealt with:

./asmcomp/selectgen.ml:798.81: [long-line] line is over 80 columns
./asmcomp/selectgen.ml:1129.81: [long-line] line is over 80 columns
./bytecomp/simplif.ml:419.81: [long-line] line is over 80 columns
./bytecomp/symtable.ml:43.3: [white-at-eol] whitespace at end of line
./bytecomp/symtable.ml:62.3: [white-at-eol] whitespace at end of line
./tools/dumpobj.ml:533.81: [long-line] line is over 80 columns
./typing/includemod.ml:167.48: [white-at-eol] whitespace at end of line
./typing/includemod.ml:168.64: [white-at-eol] whitespace at end of line
./typing/includemod.ml:319.81: [long-line] line is over 80 columns
./utils/misc.ml:180.40: [white-at-eol] whitespace at end of line
./utils/misc.mli:105.7: [white-at-eol] whitespace at end of line

damiendoligez · 2018-07-03T11:57:48Z

It is generally a bad idea to squash before the end because the commits get out of sync with the discussion and it becomes hard to make sense of them.

Drup · 2018-07-04T22:07:44Z

@dra27 All fixed, and rebased on top of trunk to take advantage of the nice typocleaning.

dra27 · 2018-07-05T07:41:17Z

Thanks for the check-typo loveliness 🙂

Drup · 2018-07-16T20:09:53Z

4.07 is released, the world cup is finished. @mshinwell, can I get a review? :)

mshinwell · 2018-07-23T08:54:52Z

Please add a comment in env.ml on the "NameMap" definition to say what "Name" means here, since we can't come up with anything better. OK to merge with that change. Thanks for taking the time to do this, it seems like a real improvement.

Drup · 2018-07-23T09:54:25Z

Done, rebased, and history somewhat cleaned up.

Drup · 2018-07-23T11:21:09Z

CI is green.

gasche · 2018-07-23T11:31:06Z

I'm a bit worried by the last commit fixing a test failure. This means that the testsuite fails if I am bisecting and end in the middle of your patchset, right? Could you maybe fixup this commit at the first point where the change is required, to maintain the invariant that intermediate states are clean?

mshinwell · 2018-07-23T11:32:33Z

I was planning to squash the whole lot; I'm not sure any intermediate state is useful here.

Drup · 2018-07-23T12:08:36Z

@mshinwell Do as you wish. :p

Drup force-pushed the tbl branch from 064c655 to bc0b7b6 Compare April 5, 2018 19:23

Octachron mentioned this pull request Apr 5, 2018

cleanup stdlib documentation makefiles #1700

Merged

Drup force-pushed the tbl branch from bc0b7b6 to 12fc7d4 Compare April 5, 2018 21:20

dra27 added the no-change-entry-needed label Apr 6, 2018

Drup force-pushed the tbl branch 2 times, most recently from 8c26653 to 9ce89b2 Compare April 6, 2018 11:50

mshinwell requested changes Apr 9, 2018

View reviewed changes

mshinwell removed the no-change-entry-needed label Apr 9, 2018

Drup force-pushed the tbl branch 2 times, most recently from ce9a7a6 to bfa53aa Compare April 10, 2018 14:02

damiendoligez assigned mshinwell May 25, 2018

Drup force-pushed the tbl branch from 547e08e to 5958159 Compare June 28, 2018 23:37

Drup force-pushed the tbl branch 2 times, most recently from b5120ed to fee2843 Compare June 29, 2018 00:26

Drup force-pushed the tbl branch from fee2843 to 5bfb6e8 Compare July 4, 2018 22:07

Drup force-pushed the tbl branch from 5bfb6e8 to 4e7d056 Compare July 4, 2018 23:01

mshinwell approved these changes Jul 23, 2018

View reviewed changes

Drup force-pushed the tbl branch from a354ec7 to 400f242 Compare July 23, 2018 09:38

Drup added 10 commits July 23, 2018 12:00

Remove Tbl in typing/

3e5ae4d

Remove Tbl in bytecomp/

746d7e7

Remove Tbl in asmcomp/

0f84ef4

Remove superfluous Sets and Maps.

57c1471

Remove Tbl from cmo files.

270a0e5

Finaly remove Tbl.

e38e4d5

Add a change entry.

ed9b1fa

Move String{Map,Set} to Misc.Stdlib.String.

eb65f54

Depend.

08e037e

Fix tests with hardcoded positions.

ad30944

Drup force-pushed the tbl branch from 400f242 to ad30944 Compare July 23, 2018 10:02

mshinwell merged commit 1be47bf into ocaml:trunk Jul 23, 2018

Drup deleted the tbl branch July 28, 2018 23:24

hhugo mentioned this pull request Mar 30, 2019

Add support for OCaml 4.08 ocsigen/js_of_ocaml#761

Merged

6 tasks

rboudrouss pushed a commit to rboudfork/ocaml that referenced this pull request Jul 12, 2026

Just some tbl things. (ocaml#1699)

d030533

Uh oh!

Conversation

Drup commented Apr 5, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Details on Tbl

Patchset

Perfs

Uh oh!

Drup commented Apr 5, 2018

Uh oh!

Octachron commented Apr 5, 2018

Uh oh!

mshinwell commented Apr 6, 2018

Uh oh!

nojb commented Apr 6, 2018

Uh oh!

dra27 commented Apr 6, 2018

Uh oh!

Drup commented Apr 6, 2018

Uh oh!

mshinwell left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Drup Apr 10, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mshinwell commented Apr 9, 2018

Uh oh!

Drup commented Apr 10, 2018

Uh oh!

Drup commented Apr 10, 2018

Uh oh!

Drup commented Jun 28, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

damiendoligez commented Jun 29, 2018

Uh oh!

gasche commented Jun 29, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Drup commented Jun 29, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

gasche commented Jun 29, 2018

Uh oh!

xavierleroy commented Jun 29, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dra27 commented Jun 30, 2018

Drup commented Apr 5, 2018 •

edited

Loading

Details on `Tbl`

Drup Apr 10, 2018 •

edited

Loading

Drup commented Jun 28, 2018 •

edited

Loading

gasche commented Jun 29, 2018 •

edited

Loading

Drup commented Jun 29, 2018 •

edited

Loading

xavierleroy commented Jun 29, 2018 •

edited

Loading