L027 bizarre log output: "Unqualified reference '\\n *' found in select"

[barrywhart]

Expected Behaviour

Observed Behaviour

Run this command:

sqlfluff fix test.sql -vvvv --exclude-rules L031 --force

On this file:

SELECT * FROM id_map AS i JOIN building_structure AS s ON i.building_id = s.building_id WHERE district_id = 3 LIMIT 5

The log output includes text:
DEBUG [L027] !! Violation Found: "Unqualified reference '\n *' found in select with more than one referenced table/view."

This is at best a bug in L027, but seeing a "reference" including newline and whitespace makes me wonder if there's a lexer or parser issue as well.

Steps to Reproduce

Dialect

Version

Include the output of sqlfluff --version along with your Python version

Configuration

Include your SQLFluff configuration here

[barrywhart]

The input file has no newlines. Thus, the newline is likely coming from a rule fix. Looking through the logs, I think I can confirm this. See this log excerpt:

INFO       0 -> 'SELECT * FROM id_map AS i JOIN building_structure AS s ON i.building_id = s.building_id WHERE district_id = 3 LIMIT 5\n'
DEBUG      [L016] !! Violation Found: 'Line is too long'
DEBUG      [L016] !! Fix Proposed: <LintFix: delete @[L:  1, P:  7] delete:' '>
DEBUG      [L016] !! Fix Proposed: <LintFix: create @[L:  1, P:  8] create:'\n    '>
DEBUG      [L016] !! Fix Proposed: <LintFix: delete @[L:  1, P:  9] delete:' '>
DEBUG      [L016] !! Fix Proposed: <LintFix: create @[L:  1, P: 10] create:'\n'>
INFO       Applying Fixes [L016]: [<LintFix: delete @[L:  1, P:  7] delete:' '>, <LintFix: create @[L:  1, P:  8] create:'\n    '>, <LintFix: delete @[L:  1, P:  9] delete:' '>, <LintFix: create @[L:  1, P: 10] create:'\n'>]
DEBUG      Matched fix against segment: <LintFix: delete @[L:  1, P:  9] delete:' '> -> <WhitespaceSegment: ([L:  1, P:  9]) ' '>
DEBUG      Matched fix against segment: <LintFix: delete @[L:  1, P:  7] delete:' '> -> <WhitespaceSegment: ([L:  1, P:  7]) ' '>
DEBUG      Matched fix against segment: <LintFix: create @[L:  1, P:  8] create:'\n    '> -> <SymbolSegment: ([L:  1, P:  8]) '*'>
DEBUG      Matched fix against segment: <LintFix: create @[L:  1, P: 10] create:'\n'> -> <KeywordSegment: ([L:  1, P: 10]) 'FROM'>
DEBUG      [L027] !! Violation Found: "Unqualified reference '\\n    *' found in select with more than one referenced table/view."
DEBUG      [L027] !! Violation Found: "Unqualified reference '\\n    *' found in select with more than one referenced table/view."
DEBUG      [L027] !! Violation Found: "Unqualified reference 'district_id' found in select with more than one referenced table/view."

So the L016 fixes are probably causing something weird.

[barrywhart]

I examined the segments involved at the time L027 reports the issue.

(Pdb) !r
<WildcardIdentifierSegment: ([L:  1, P:  8])>
(Pdb) !r.segments
(<NewlineSegment: ([L:  1, P:  8]) '\n'>, <WhitespaceSegment: ([L:  1, P:  8]) '    '>, <SymbolSegment: ([L:  1, P:  8]) '*'>)

If I modify test.sql so L016 didn't need to split the line, e.g.

Now L027 still detects an issue, but the WildcardIdentifierSegment only has one child.

(Pdb) !r
<WildcardIdentifierSegment: ([L:  2, P:  5])>
(Pdb) !r.segments
(<SymbolSegment: ([L:  2, P:  5]) '*'>,)

[barrywhart]

This issue looks like another occurrence of #1304.

[tunetheweb]

Tested this on latest main and don't get that any more.

Do get a few others, but think it's right they are there:

% sqlfluff fix test.sql -vvvv --exclude-rules L031 --force | grep "ound in select with more than one refere"
DEBUG      [L027] !! Violation Found: "Unqualified reference '*' found in select with more than one referenced table/view."                                                                                                                             
DEBUG      [L027] !! Violation Found: "Unqualified reference 'district_id' found in select with more than one referenced table/view."