Cannot access "non-node" arguments in drop down widgets evaluation

[JaroslavTulach]

There has to be a balance between amount of cached values and optimal memory usage. As such the PR

• Access non-self arguments when widget annotations are computed #9410

added following reasoning and provides access only to argument values already assigned to some other node. @jdunkerley points out that's not enough:

make_format_selector table display=Display.Always cache=Nothing =
    columns = cache.if_not_nothing <| cache "columns"

  @columns (Widget_Helpers.make_column_name_multi_selector add_regex=True)
    @locale Locale.default_widget
    @format Widget_Helpers.make_format_selector
    format : Vector (Text | Integer | Regex) | Text | Integer | Regex -> Text | Date_Time_Formatter | Column -> Locale -> Boolean -> Problem_Behavior -> Table ! Date_Time_Format_Parse_Error | Illegal_Argument
    format self columns format:(Text | Date_Time_Formatter | Column)="" locale=Locale.default error_on_missing_columns=True on_problems=Report_Warning =

Workaround:

Are you calling operator.format [ "x", "y" ]?
Try to move [ ... ] into its own node.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-05-21):

Progress: - playing with `Widget: #9941

• addressing review comments: Register all polyglot java import classes for refective access #9997 (comment)
• perf issues: Analyze and improve the performance of Import/export resolution #9236 (comment)
• reporting PoC LS as library: PoC: Package language-server as a library #10010
• planning work & meeting & standup It should be finished by 2024-05-28.

[JaroslavTulach]

I've created a small project to play with the drop down arguments. Download as AccessArgumentsDropDown.zip and unzip. The code in question is:

from Standard.Base import all
from Standard.Base.Metadata import all
import Standard.Visualization

type T
    @b for_b
    as_list a b = [a, b]

for_b s cache=Nothing =
    me = [ "self", s ]
    a = ["a", cache "a"]
    b = ["b", cache "b"]
    make [1, "dva", me, a, b]

make values:Vector =
    make_option value = Choice.Option value.to_text value.pretty
    dbg values.to_text
    Widget.Single_Choice (values.map make_option) Nothing Display.Always

main =


    integer1 = 11
    node1 = T.as_list 436 1
    [node1, _]

foreign js dbg info = """
    debugger;

The first thing to notice is that evaluation of the @b widget content happens at UUID 83b40fc8-1b02-418e-8e0d-80f153bc2504 (index 496, size 9) which represents T.as_list - however when this node is evaluated, its arguments 436 and 1 aren't yet computed. There is also d872a8e8-47ea-4c0c-b3ef-a20df2ab4c3b which represents the whole T.as_list 436 1. If we want the drop down visualization to access the arguments, shouldn't we evaluate the visualization for the whole expression, @Frizi and @farmaazon?

[JaroslavTulach]

@jdunkerley, @marthasharkey - if you want to experiment with accessing arguments like 436, try it with this patch:

diff --git engine/runtime-instrument-common/src/main/java/org/enso/interpreter/instrument/RuntimeCache.java engine/runtime-instrument-common/src/main/java/org/enso/interpreter/instrument/RuntimeCache.java
index da2f8e2447..985a02d943 100644
--- engine/runtime-instrument-common/src/main/java/org/enso/interpreter/instrument/RuntimeCache.java
+++ engine/runtime-instrument-common/src/main/java/org/enso/interpreter/instrument/RuntimeCache.java
@@ -38,14 +38,8 @@ public final class RuntimeCache implements java.util.function.Function<String, O
       valuesToKeys.put(value, key);
       return true;
     } else {
-      var otherId = valuesToKeys.get(value);
-      var otherValue = cache.get(otherId);
-      if (otherValue != null && otherValue.get() == value) {
-        // if the value is already cached for another UUID, cache it
-        // for this expression UUID as well
-        var ref = new WeakReference<>(value);
-        expressions.put(key, ref);
-      }
+      var ref = new WeakReference<>(value);
+      expressions.put(key, ref);
       return false;
     }
   }

it does seem to work to some extend (values may be outdated, or missing from time to time). It should be good enough for testing.

[farmaazon]

The first thing to notice is that evaluation of the @b widget content happens at UUID 83b40fc8-1b02-418e-8e0d-80f153bc2504 (index 496, size 9) which represents T.as_list

By evaluation you mean on what value get_widget_json is called? Then it's even stranger, because according to our code it should be called on self argument, that is T in this case.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-05-22):

Progress: - small Widget repro: https://discord.com/channels/@me/948519891961544714/1242708122607550464

• design evaluation for widgets: Cannot access "non-node" arguments in drop down widgets evaluation #9941 (comment)
• addressing review comments: Register all polyglot java import classes for refective access #9997
• review of Hubert's: Extract mutable builder from ModuleScope #9914 (review)
• meeting Pavel
• meeting James: rewrite table representation (to Truffle?)
• reacting to comments: aaa5e9e It should be finished by 2024-05-28.

Discord

Discord - A New Way to Chat with Friends & Communities

Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.

[JaroslavTulach]

The first thing to notice is that evaluation of the @b widget content happens at UUID 83b40fc8-1b02-418e-8e0d-80f153bc2504 (index 496, size 9) which represents T.as_list

By evaluation you mean on what value get_widget_json is called?

Yes. get_widget_json calls into @b annotation.

Then it's even stranger, because according to our code it should be called on self argument, that is T in this case.

You are right the [{"index":{"value":496},"size":{"value":1}},"83b40fc8-1b02-418e-8e0d-80f153bc2504"] node represents T - I misread the location due to strange formatting of the metadata in my text editor.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-05-23):

Progress: - ms in log merged : #10016

• Enso Library Feature merged: Register all polyglot java import classes for refective access #9997
• reporting subsequent work: Preload library caches into native-image heap to speed start up #10044
• widgets discussion with Pawel & Hubert & at standup
• "storage benchmark" proposal: Benchmark Truffle "add" for arrow-language and Column.+ #10056 It should be finished by 2024-05-28.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for the last Friday (2024-05-24):

Progress: - reactive RuntimeCache PR: #10065

• debugging and recording video: Reactive RuntimeCache #10065 (comment)
• reviewing persistance fixes by Radek: https://discord.com/channels/@me/964018046375190528/1243541832156319896
• planning, standups It should be finished by 2024-05-28.

Discord

Discord - A New Way to Chat with Friends & Communities

Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-05-27):

Progress: - Reactive Cache is in: #10065

• pesistance fix: https://discord.com/channels/@me/964018046375190528/1244553530661015622
• review, not really in favor of attach-debugger: --attach-debugger to start Language Server with debugger from project manager #10062 (review)
• ModuleScope.Builder note: https://github.com/enso-org/enso/pull/9914/files#r1615844358
• review private method: Implement private methods #10060 (review)
• debugging suspended default arguments: Suspended default arguments cannot access previous arguments (including self) #8937 (comment) It should be finished by 2024-05-28.

Discord

Discord - A New Way to Chat with Friends & Communities

Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.

GitHub

WIP: Extract mutable builder from `ModuleScope` by hubertp · Pull Request #9914 · enso-org/enso

Pull Request Description Refactored mutable parts of ModuleScope into builder to make it easier to reduce unnecessary locks. Important Notes

Checklist
Please ensure that the following checklist ha...