REPL: Don't search for ?( completions when hinting

[IanButterworth]

Fixes #56272

[giordano]

Do I understand correctly this is only ignoring a line starting with a string?

[IanButterworth]

Yeah I think this needs fixing. How does the below proposal look?

(\w+\.|)\?\((.*)$ previous
^(?!\")(?:\w+\.|)\?\((.*)$ current
^\s*(?!\")(?:\w+\.|)\?\(\s*(.*)\s*$ proposed here

julia> regex = r"^\s*(?!\")(?:\w+\.|)\?\(\s*(.*)\s*$"
r"^\s*(?!\")(?:\w+\.|)\?\(\s*(.*)\s*$"

julia> match(regex, "?(")
RegexMatch("?(", 1="")

julia> match(regex, "\"?(\"")

julia> match(regex, "\"\"\"?(\"\"\"")

julia> match(regex, "\"?(\"")

julia> match(regex, " \"?(\"")

julia> match(regex, " \" ?(\"")

julia> match(regex, " ?(\"")
RegexMatch(" ?(\"", 1="\"")

julia> match(regex, " ?( ")
RegexMatch(" ?( ", 1="")

julia> match(regex, " ?(")
RegexMatch(" ?(", 1="")

ChatGPT summary:

Let's break down the regular expression ^\s*(?!\")(?:\w+\.|)\?\(\s*(.*)\s*$ step by step:

1. ^:

• This asserts the start of the string. The pattern must begin matching from the start.

2. \s*:

• \s matches any whitespace character (spaces, tabs, etc.).
• * means "zero or more" occurrences of the preceding element, so this part matches zero or more whitespace characters at the start of the string.

3. (?!"):

• This is a negative lookahead. It ensures that the next character is not a double quote (").
• It checks if the string does not begin with a " without consuming any characters. If the string starts with a double quote, the match will fail.

4. (?:\w+.|):

• (?: ... ): This is a non-capturing group, meaning it groups part of the pattern but does not capture it for backreferences.
• \w+: This matches one or more word characters (letters, digits, or underscores).
• \.: This matches a literal dot (.).
• |: This is the "OR" operator, meaning the non-capturing group matches either \w+. (a word followed by a dot) or nothing (i.e., an empty string).

This means the string may optionally have a prefix like word. before the ?(.

5. ?(:

• This matches the literal sequence ?(. It looks for the exact string ?(.

6. \s*:

• This matches zero or more whitespace characters after the ?(. It allows for spaces between ?( and the content that follows.

7. (.*):

• This is a capturing group that matches any characters (except newlines) 0 or more times.
• The content matched by this group is captured, and it can be referenced or used later.
• This captures the content inside the parentheses following ?(, including an empty string.

8. \s*:

• This matches zero or more whitespace characters at the end of the string, allowing for trailing spaces after the content.

9. $:

• This asserts the end of the string. The pattern must match up to the end.

Summary

The regular expression ^\s*(?!\")(?:\w+\.|)\?\(\s*(.*)\s*$ matches strings that:

• Can start with optional whitespace.
• Do not start with a double quote (").
• Optionally have a word (letters, digits, or underscores) followed by a dot (e.g., word.).
• Contain the literal sequence ?(, possibly surrounded by spaces.
• Capture everything after the ?(, including an empty string.
• Can have optional whitespace after the captured content.
• Must match from the beginning (^) to the end ($) of the string.

Examples that would match:

1. " ?(" (matches and captures empty content).
2. " ?( some content ) " (matches and captures "some content").
3. " example.?(content)" (matches and captures "content").
4. " ?( " (matches and captures empty content with spaces).

[IanButterworth]

@Seelengrab just saw your comment in the issue, if you can help resolve this here

[mbauman]

I mean, you can only determine "inside string" with a complete parser, I believe. I don't think we really need to fiddle with the regex with the !hint restriction that's now here.

I had been thinking that any additional restrictions to the regex would be complementary, but both the current and proposed would limit suggestions in cases that currently work, like f(?(1,2)tab.

[IanButterworth]

ok. I'll revert the regex change and get the painful part fixed with the !hint

[mbauman]

Is there another way we can avoid these kinds of completions for the end-of-line suggestions? This seems helpful, but not quite complete.

[KristofferC]

Is there another way we can avoid these kinds of completions for the end-of-line suggestions

+1, just disable for inline completions.

[mbauman]

Easiest to review ignoring whitespace: https://github.com/JuliaLang/julia/pull/56278/files?w=1. This looks good to me, thanks @IanButterworth!