Seal heredoc renders so nothing packs onto their terminator by JPDuchesne · Pull Request #17 · rspockframework/ast-transform

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne · 2026-08-03T14:45:57Z

Summary

FixesLine-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16: when Unparser falls back to <<-HEREDOC emission (a dstr failing its quoted round-trip — e.g. interpolation-heavy strings carrying newlines), the line-aligned emitter packed displaced statements and container closers onto the terminator line with ; , unterminating the heredoc and swallowing the rest of the file.
StatementRenderer#line_terminal? decides whether a render's last line can be extended, by re-parse verification (the same style compress_to_single_line already uses) behind a cheap << textual gate. The probe chomps first: heredoc renders end with a trailing newline, and the question is about the last line, where packing would land.
Layout#place accepts the verdict as a seal: property, keeping the layering intact (Layout stays Ruby-free — "this last line is unextendable" is a caller-declared property). A pack onto a sealed line opens a fresh line instead: alignment degrades by one line, validity is preserved. The seal covers exactly one pack; padding to a line ahead is unaffected.
Both statement renders and container openers seal (an opener can carry a heredoc argument too). A false-positive seal costs one line of alignment, never correctness.

Test plan

End-to-end regression in line_aligned_emitter_test.rb: the Line-aligned emission places statements after a heredoc terminator, producing invalid Ruby #16-shaped source (heredoc-fallback render displacing following statements) now compiles, terminator alone on its line — failed with the issue's exact SyntaxError before the fix
Layout seal semantics: sealed pack opens a fresh line, seal covers one pack, padding unaffected
StatementRenderer#line_terminal?: heredoc render flagged; plain renders, shovel <<, quoted "<<-EOS", and multi-line quoted strings pass
Full suite (112 tests) and rubocop green

Made with Cursor

Unparser legitimately falls back to <<-HEREDOC emission when a dstr fails its quoted-string round-trip; the line-aligned emitter then treated that render like any other statement and `;`-packed displaced statements (and container closers) onto the terminator line, unterminating the heredoc and producing invalid Ruby (#16). StatementRenderer#line_terminal? probes whether a render's last line can be extended (re-parse verification, the same style as compress_to_single_line, behind a cheap `<<` gate), and Layout#place accepts the result as a `seal:` property: a pack onto a sealed line opens a fresh line instead — alignment degrades by one line, validity is preserved. Statements and container openers both seal; a false-positive seal costs alignment, never correctness. Closes#16 Co-authored-by: Cursor <cursoragent@cursor.com>

codecov · 2026-08-03T14:47:42Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

JPDuchesne merged commit 6ab5d2d into mainAug 3, 2026
5 checks passed

JPDuchesne deleted the jpd/16-heredoc-line-terminal branch August 3, 2026 14:48

This was referenced Aug 3, 2026
Release 3.1.1 #18
Merged
Bump ast_transform to 3.1.1 and drop the heredoc-emission workaround d3mlabs/dev#88
Merged