Let EnsoMultiValue.to_text delegate to first type to_text

[JaroslavTulach]

Based on to_text discussion, but overall question is: what should EnsoMultiValue do with Any methods? If we delegate to_text to the first type, then we should probably delegate all of them. @radeusgd originally wrote:

IMO to_text of first type is best. It does hide the 'other values' and that is not ideal, but given the kind of use cases we want to use multi-values for (e.g. the Column & Integer_Column), I think that delegating to the first type is the best we can do to keep this sensible.

In that regard, perhaps to_text shouldn't be any different than e.g. to_js_object or to_defautl_visualization_data that we use to visualize the values. These will also delegate to the first value.

Originally posted by @radeusgd in #11600 (comment)

[JaroslavTulach]

As a side note: Any.to should probably use EnsoMultiValue as self: #11897 (comment)

Update on Jan 24, 2025: Another example where EnsoMultiValue should be send as self is here https://github.com/enso-org/enso/pull/12120/files#r1928099736 - again needed for dispatch of Any instance methods.

[radeusgd]

As a side note: Any.to should probably use EnsoMultiValue as self: #11897 (comment)

Update on Jan 24, 2025: Another example where EnsoMultiValue should be send as self is here https://github.com/enso-org/enso/pull/12120/files#r1928099736 - again needed for dispatch of Any instance methods.

Indeed it would be good if methods like catch or throw_on_warning did not lose the intersection types. It is not intuitive that they would be lost (I understand why it happens but writing the code it is not an intuitive behaviour) - it requires the user to think hard whether their code may be getting an intersection type and do workarounds to avoid losing it. That is definitely not desirable - I think the intersection types should be as transparent as possible.

GitHub

Refine single-value column to treat it as that single value by radeusgd · Pull Request #12120 · enso-org/enso

Pull Request Description

First experiments with intersection types as refinements to our Table/Column
Closes #12095

Important Notes

Checklist
Please ensure that the following checklist has been ...

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2025-02-11):

Progress: .

• ready for review: Dispatch of to_text, is_a & co. on intersection types #12192
• Trying pyarrow: Investigate Parquet/Avro usage in Enso #8772 (comment)
• review ydoc-server/test: Enable ydoc-server tests #12252 (review)
• evaluating benchmarks regression: WarningBenchmarks 5000% regressions on 2025-02-11 #12258 (comment) It should be finished by 2025-02-14.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2025-02-12):

Progress: .

• making up to date: Removing State from method signatures #12233
• bisecting: WarningBenchmarks 5000% regressions on 2025-02-11 #12258 (comment)
• benchmarking State: Removing State from method signatures #12233 (comment)
• merged Dispatch of to_text, is_a & co. on intersection types #12192
• analyzing IGV graphs for State PR
• running benchmarks again: Removing State from method signatures #12233 (comment)
• (standup) meetings It should be finished by 2025-02-14.

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2025-02-13):

Progress: .

• reviewing Frame_Hider: https://github.com/enso-org/enso/pull/12260/files#r1953801227
• comparing IGV graphs for VectorBenchmarks.averageOverSliceWrapped100
• libs standup
• IGV for State It should be finished by 2025-02-14.

GitHub

Make test failure source locations lazy by GregoryTravis · Pull Request #12260 · enso-org/enso

Replaces eager calls to Test.get_source_location to lazy calls to Frame_Hider.get_top_stack_frame, resulting in a 2x speedup for Base_Tests. Test utilities wrap their contents in Frame_Hider.hide, ...