if then else and chained block syntax

[JaroslavTulach]

Assuming that

    main =
        pairs = (if t then a else b) . to_vector . sort

and

    main =
        pairs = if t then a else b
            . to_vector
            . sort

shall be the same. They are not right now on 2023.2.1-nightly.2023.12.7

[somebody1234]

@JaroslavTulach i think you messed up the link?

[JaroslavTulach]

@kazcw what do you think? As far as I can see in debugger the chained code is equivalent to b . to_vector . sort - e.g. the chain is only applied to the b and not to the result of the whole if_then_else.

[JaroslavTulach]

There are three tests. Two work, but testIfAndBlockSyntax fails. Is that correct or wrong, @kazcw?

[kazcw]

I'm looking for a consistent rule that would make this work. Here's a candidate:

The body of an else segment is either:

• The rest of the line (if non-empty)
• A following child block (if there is no expression following the else on the same line)

This is a change from current behavior, in which if the else has in-line content, any following child block belongs to the else branch.

This would also affect other block types, e.g. this:

    main =
        value = if t then f1 else f2
            y

Would be equivalent to this:

    main =
        value = (if t then f1 else f2)
            y

IMO it minimizes confusion to treat all block types the same way in this regard.

What do you think, @JaroslavTulach? If we shall do it, I can use the AST-comparison script to check for any existing code relying on the old behavior.

[JaroslavTulach]

The rest of the line (if non-empty)
A following child block (if there is no expression following the else on the same line)

+1

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-02-21):

Progress: - rust parser: https://github.com/enso-org/enso/pull/8671/files#r1497057518

• review logging: Unexpected warning when running a standalone script #9114 (review)
• review SemVer: Drop inefficient SemVer implementation #9089 (review)
• discussing atom_with_hole
• fixing Polyglot_Spec: Including missing Polyglot_Spec in all the tests #9126
• Extension method for Java classes PR: Extension methods for Java classes #9129
• meetings and discussions It should be finished by 2024-02-24.

Next Day: Rust parser, Extension methods for Java

GitHub

Chained if_then_else application change by JaroslavTulach · Pull Request #8671 · enso-org/enso

Pull Request Description Attempt to fix #8489 by

adding test describing current behavior: c5bef79
another test showing desired Group-like behavior: 1c722f4
changing the first test to behave like t...

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-02-22):

Progress: - chained if_then_else re-review: #8671 (review)

• supporting static Java extension methods
• and imports: Extension methods for Java classes #9129 (comment)
• fighting with "static syntax" invocation
• describing future direction: Extension methods for Java classes #9129 (comment)
• enso4igv bugfix: d74ee24 It should be finished by 2024-02-24.

Next Day: Extension methods for Java

[enso-bot]

Jaroslav Tulach reports a new STANDUP for yesterday (2024-02-23):

Progress: - bookclubbing

• emails, standups, discussions
• reported ifBench6 slowdown: IfThenElse IR to avoid 3x slower IfVsCaseBenchmarks_ifBench6In #9165
• reported collections slowdown: 10x slowdown in Collections_list_meta_fold #9166
• Using Enso as GraalVM EE Isolate. It works! It should be finished by 2024-02-24.

Next Day: Extension methods for Java