Enable canonical .eo format check in the eo-runtime build by yegor256 · Pull Request #5579 · objectionary/eo

yegor256 · 2026-07-16T14:03:48Z

This wires the eo-maven-plugin format goal into eo-runtime's compile execution and reformats all 133 .eo sources to the canonical layout the printer produces. With the goal in check mode (autoFix off), any source that drifts from canonical form now fails the build.

The point is to make formatting a hard gate rather than a convention. This is the eo-runtime slice of #5577, which proposes the same check for every module that carries .eo sources.

Verified locally: a clean build reports "All 139 EO source(s) are formatted canonically", and the full test suite passes (1551 tests, 0 failures). Worth knowing for review: the printer's canonical output will shift slightly once #5567 and #5570 land (compact foo ++> name and short +alias form), so a follow-up reformat will be needed then; and the remaining modules named in #5577 still need the same wiring.

Wire the eo-maven-plugin `format` goal into the compile execution so a non-canonical `.eo` source fails the build, and reformat all existing sources to the canonical layout produced by the printer so the check passes.

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

github-actions · 2026-07-16T14:06:04Z

⚠️ Benchmark Comparison Unavailable

Unfortunately, one of the benchmarks is missing, and we couldn't generate a performance comparison report.

Please ensure that both the base and PR benchmark results are available for analysis.

Add a per-symbol penalty (one point per character) and make the bracket penalty progressive: the k-th parenthesis on a line costs k times its weight, so stacking parentheses on one line is discouraged superlinearly. Retune the weights: indent 3 to 5, bracket 7 to 15, excess 1 to 3. Together these push the printer away from long parenthesised inline forms toward flatter vertical layouts. Reformat the eo-runtime sources to the new canonical layout and update the PenaltyTest expectations and two print-pack goldens (compares-bool-to-bytes and times).

The single-parenthesis case (default weight) is already covered by the progressive-parentheses test (two parens = 3x the weight, which also pins the default weight) and by the overridden-weight test. Removing it keeps the class under the method-count limit and drops the duplicated string literal below the threshold, so qulice passes.

Master's #5567 fix (PR #5582) added the compact `foo ++> name` test shorthand to the parser and printer. Re-run the format goal so every eo-runtime source adopts that shorthand together with the retuned penalty weights, keeping the sources canonical under the merged printer.

# Conflicts: # eo-printer/src/test/resources/org/eolang/printer/print-packs/yaml/times.yaml

Re-run the format goal after merging master's printer fixes for #5583 (atom signatures keep their Q. root) and #5584 (data-receiver dispatch prints in suffix form). Sources now round-trip to valid XMIR and adopt the suffix dispatch, staying canonical under the merged printer.

Regenerate the canonical .eo sources now that the printer bugs (#5581, #5583, #5584, #5589, #5591) are fixed on master. The re-format is idempotent and the full eo-runtime suite passes (1551 tests, 0 failures, 0 errors).

Cheaper indentation makes the layout search prefer vertical renderings over deeply parenthesized inline ones. Update the two round-trip golden packs (dataless, empty-string) and the two PenaltyTest expectations that pinned the old default.

Regenerate the canonical .eo sources under the lowered indentation penalty. The re-format is idempotent and the full eo-runtime suite passes (1551 tests, 0 failures, 0 errors).

Re-run the format goal now that the printer withholds the hybrid inline-phi collapse when any decoratee descendant is named (#5604), so file.eo, tuple/reducedi.eo and while.eo re-parse cleanly. The output is idempotent across two passes and the full eo-runtime suite passes.

yegor256 · 2026-07-17T07:34:17Z

@maxonfjvipon take a look again (ignore the xcop failure for now)

maxonfjvipon · 2026-07-16T14:08:49Z

      slice-byte
        tup.with (bts.slice index 1)
-        (index.plus 1).as-number
+        as-number. (index.plus 1)


@yegor256 why?

maxonfjvipon · 2026-07-17T08:08:44Z

-        123
-        42
-      123
+  eq. (if. (5.eq 5) 123 42) 123 ++> tests-forks-on-true-condition


@yegor256 that's not quite readable

maxonfjvipon · 2026-07-17T08:09:12Z

-          eq. > @
-            x
-            42
+        ^.^.^.x.eq 42 > [] > @


@yegor256 too much noisy rhos

maxonfjvipon · 2026-07-17T08:10:20Z

-  [start len] > slice /bytes
-    ? > cant-slice /{string}
-
+  [start len cant-slice] > slice /bytes


@yegor256 we loose /{string} here

@maxonfjvipon bug: #5614

maxonfjvipon · 2026-07-17T08:11:35Z

-      plus.
-        timeb.time.times 1000
-        timeb.millitm
+    sm.os.is-windows.if


@yegor256 if. is much more readable than .if

maxonfjvipon · 2026-07-17T08:14:13Z

 +unlint mandatory-package

-2.7182818284590452354 > e
+2.718281828459045 > e


@yegor256 precision loosing

@maxonfjvipon it was over-precise: #5613

maxonfjvipon · 2026-07-17T08:16:20Z

          *
        if.
-          contains prev.hashes hash
+          tuple.contains prev.hashes hash


@yegor256 unnecessary tuple

maxonfjvipon · 2026-07-17T08:17:16Z

-          getenv "TMP" > tmp!
-          getenv "TMPDIR" > tmpdir!
-          getenv "USERPROFILE" > userprofile!
+          as-bytes. > temp


@yegor256 ! is lost

@maxonfjvipon bug: #5618

maxonfjvipon · 2026-07-17T08:18:58Z

-
+  seq > @
+    * (output.write text) true
+  sm.os.is-windows.if windows-output posix-output > output


@yegor256 unreadable

maxonfjvipon · 2026-07-17T08:19:35Z


-  ++> throws-on-taking-length-of-incomplete-n4-byte-character
-    (string F0-9F-98).length > @
+  "�".length ++> throws-on-taking-length-of-incomplete-n4-byte-character


@yegor256 ???

# Conflicts: # eo-runtime/src/main/eo/bytes.eo # eo-runtime/src/main/eo/bytes/as-i16.eo # eo-runtime/src/main/eo/bytes/as-i32.eo # eo-runtime/src/main/eo/bytes/as-i64.eo # eo-runtime/src/main/eo/bytes/as-number.eo # eo-runtime/src/main/eo/directory.eo # eo-runtime/src/main/eo/file.eo # eo-runtime/src/main/eo/i16.eo # eo-runtime/src/main/eo/i32.eo # eo-runtime/src/main/eo/i64.eo # eo-runtime/src/main/eo/map.eo # eo-runtime/src/main/eo/number.eo # eo-runtime/src/main/eo/number/mod.eo # eo-runtime/src/main/eo/string.eo # eo-runtime/src/main/eo/string/as-number.eo # eo-runtime/src/main/eo/string/at.eo # eo-runtime/src/main/eo/string/nsplit.eo # eo-runtime/src/main/eo/string/printf.eo # eo-runtime/src/main/eo/string/regex.eo # eo-runtime/src/main/eo/string/repeated.eo # eo-runtime/src/main/eo/string/scanf.eo # eo-runtime/src/main/eo/switch.eo # eo-runtime/src/main/eo/tuple.eo # eo-runtime/src/main/eo/tuple/maximum.eo # eo-runtime/src/main/eo/tuple/minimum.eo # eo-runtime/src/main/eo/tuple/range-of-numbers.eo

Re-run the format goal over eo-runtime after merging the printer round-trip fixes from master (#5620 lossy UTF-8 decode, #5622 and #5624 trailing-star hybrid). The output is now idempotent (a second format pass reports no divergence) and provably lossless: every reformatted source parses to the same object tree as before, modulo line numbers and position-derived auto-names.

The penalty function now charges APPLICATION (7 points by default) for every space on a line beyond its leading indentation, so bracket- and space-heavy horizontal forms give way to vertical ones. All print-packs pin the new weight and a handful of goldens are reconciled to the new layout.

Re-run the canonical formatter over all .eo sources now that the printer charges the APPLICATION penalty, breaking bracket- and space-heavy horizontal lines onto vertical layouts. 75 of 139 files changed. The reformat is idempotent, every source parses, and all XMIR trees are identical to before, so the change is purely cosmetic.

Pull master's #5650: suffixed() now rebuilds any reversed dispatch whose receiver has a one-line form as "(receiver).method", not only lone-data receivers. The merge keeps 5577's penalty weights, so this suffix-glue is generated and wins wherever it beats the reversed vertical form. Regenerate the seven print-pack fixtures whose pinned output shifted to the resulting hybrid: "(x).method" with the remaining arguments kept vertical. All 141 eo-printer tests pass. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Re-canonicalize the runtime .eo under master's #5650, just merged in. A reversed dispatch whose receiver has a one-line form now prints as "(receiver).method" wherever that beats the vertical reversed form, collapsing many two- and three-line test attributes onto one line (e.g. "(true.eq 42).not ++> tests-..."). Idempotent and lossless: all 97 changed files parse to the same object tree as before. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The descriptive comment blocks restate what each pack's origin, penalties and printed fields already show. Remove them, keeping the SPDX header and the yamllint line-length directive. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Drop the printer's default INDENT weight from 3 to 2 so vertical layout costs less and wins more often over compact bracketed dispatch. Adjust the two penalty assertions that read the INDENT default (five levels 15->10, two levels 6->4). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Re-canonicalize the runtime .eo now that vertical layout is cheaper. Many compact "(receiver).method" one-liners give way to the vertical reversed dispatch again, since the saved indentation no longer offsets the bracket. Idempotent and lossless: all 80 changed files parse to the same object tree as before. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The penalty-based printer charged its per-space weight flatly, once per space on a line, so each extra argument cost the same constant increment no matter how long the application already was. Nothing pushed the printer, or the author, away from a sprawling application. Now the genuine argument-applying spaces on a line pay a super-linear surcharge: r such spaces cost r squared, rather than r, times the weight, so longer applications grow super-linearly more expensive and the printer favours shorter ones (for the three arguments in "foo bar 42 44" the weight is effectively multiplied by three). Spaces that merely bind a name (">", ">>", "++>") or separate the void attributes inside a formation's "[..]" head are not applications, so they keep their flat per-space charge and are left untouched: "foo > [] > bar" is unchanged. Renamed the PenaltyKey.APPLICATION weight to PenaltyKey.SPACE, since it is charged for every space on a line, not only for argument applications, and the print-pack weight blocks with it. Regenerated the affected eo-printer print-pack goldens. Scoped to eo-printer; reformatting eo-runtime sources is a separate concern and is not included here. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01KL51NxsYxxo34h6s7iVByZ

#5653: scale the per-space penalty super-linearly by application length

Re-canonicalize eo-runtime sources with the eo-printer from PR #5654, whose per-space penalty now grows as the square of an application's length. Sprawling multi-argument applications wrap into vertical form. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Bring in #5655, which adds a PHI(15) penalty steering the printer away from explicit @ renderings. Combined with this branch's richer model (progressive BRACKET, super-linear SPACE, INDENT 2): kept the progressive bracket cost and super-linear space surcharge, added the phi charge alongside them, and refreshed the PenaltyTest expectations and the two worked Javadoc examples to the merged weights. Updated the multiline-string print-pack golden, whose preferred rendering now drops the @ suffix now that phi is charged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Re-canonicalize eo-runtime sources with the merged eo-printer, whose new PHI penalty from #5655 charges explicit @ renderings. Blocks that bind their phi now fold into shorter phi-free suffix forms. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

`Penalty.binding()` recognised `>`, `>>` and `++>` as name-binding markers but omitted the throwing-test marker `-->`. As a result the two spaces flanking `-->` were charged as argument applications and paid the super-linear SPACE surcharge, so the phi-free shorthand `mod 5 0 --> on-mod-by-zero` scored far higher than its `++>` twin and the layout search kept the verbose `--> name / body > @` rendering with an explicit `@`. Add `"-->".equals(token)` to the disjunction so the marker is treated like `++>`, restoring the collapse (this is the regression #5653 introduced for `-->` when it added the surcharge, the untreated twin of #5567). Update the `throws-test-attribute` print-pack golden back to the collapsed form, add a `PenaltyTest` case, and reformat the eo-runtime `.eo` sources whose `-->` tests now render canonically in the compact shorthand. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RjwDHuWPVtWq2kv4VrCFyD

fix(#5657): treat throwing-test marker `-->` as a name binding

The print-packs are shared (via a symlink) between eo-printer's XmirTest, which renders them under each pack's pinned weights, and MjPrintTest, which rendered them under the printer's default weights. After the branch lowered the default INDENT weight, 16 packs diverged and MjPrintTest broke in CI. Give MjPrint the same penalty parameters MjFormat already exposes and feed each pack's pinned weights to the mojo, so both tests lay a pack out under the weights it declares.

Resolve the bracket-penalty conflict by adopting master's depth-weighted brackets() from PR #5660 (each parenthesis costs its nesting depth plus one) into this branch's richer six-term penalty model, charged linearly, and dropping the branch's per-line n(n+1)/2 count-progression. For nested brackets the two coincide, so no runtime .eo formatting changed; the one affected print-pack, inline-phi-overflow, now renders vertically because the SPACE surcharge outweighs the deep one-liner. Also clear two pre-existing jtcop violations that block the qulice job: rename the throwing-marker test to drop the word "test", and move the weight-override tests out of the orphan PenaltyOverridesTest into a new PenaltyKeyTest that maps one-to-one to PenaltyKey.

yegor256 · 2026-07-20T14:00:12Z

@maxonfjvipon take another look

maxonfjvipon · 2026-07-20T14:19:01Z

  ++> tests-iterating-tuple-with-while-using-internal-iterator
    data.eq arr.length > @
-    * 1 1 1 1 > arr
+    * > arr


@yegor256 why no horizontal tuple here?

maxonfjvipon · 2026-07-20T14:19:19Z

          [iter] >>
-            if. > @
-              max.eq 0
+            (^.^.max.eq 0).if > @


@yegor256 still explicit rhos

maxonfjvipon · 2026-07-20T14:21:17Z

-  --> on-empty-switch
-    switch * > @
+  switch --> on-empty-switch
+    *


@yegor256 why no horizontal only-phi here?

switch * --> on-empty-switch

maxonfjvipon · 2026-07-20T14:21:40Z

      true
-  (rec-case cases).@ > found
-
+  @. > found


@yegor256 still not sure about this

maxonfjvipon · 2026-07-20T14:23:45Z

 +unlint mandatory-package

-number 7F-F8-00-00-00-00-00-00 > nan
+number > nan


@yegor256 not sure what's better here

maxonfjvipon · 2026-07-20T14:24:17Z

-                    tempdir.eq empty
-                    "/tmp"
-                    tempdir
+            (tmpdir.eq empty).if


@yegor256 this ).if is not intuitive

maxonfjvipon · 2026-07-20T14:26:25Z

          read source length
-
-      [offset length] > read /bytes
-        ? > cant-read /{string}


@yegor256 still loosing important atom info

The runtime was reformatted before #5614 taught the printer to keep the /{...} forma-list on an atom's error-branch void, so malloc's read atom had been folded to [offset length cant-read], dropping the cant-read /{string} type. Reformatting the master source with the current printer restores it; the other 138 runtime sources come out byte-identical to their committed form.

With #5664 merged from master, the printer now collapses multi-hop ^.^. parent chains to bare names when scope resolution re-derives the hops. Re-reformatting the master sources with the updated printer strips those redundant hops from 16 runtime files (while.eo and others). The rewrite is lossless — every file parses to the same XMIR as master — and idempotent.

# Conflicts: # eo-printer/src/main/java/org/eolang/printer/Penalty.java # eo-printer/src/main/java/org/eolang/printer/PenaltyKey.java # eo-printer/src/test/java/org/eolang/printer/PenaltyTest.java # eo-printer/src/test/resources/org/eolang/printer/print-packs/yaml/nested-if-inline.yaml # eo-printer/src/test/resources/org/eolang/printer/print-packs/yaml/seq-star-idiom.yaml # eo-printer/src/test/resources/org/eolang/printer/print-packs/yaml/void-local-handle.yaml

After merging master's #5666, the printer charges a penalty for suffix `if` (`foo.if`), so the format goal now renders every `if` in prefix form. Re-ran the goal over master's sources; output is idempotent and parses to identical XMIR.

Merging master's #5666 added `ifs` and `identifier`, which together with the branch's space helpers pushed Penalty to twelve methods and tripped PMD TooManyMethods. Inline `overflow` into `points` and `identifier` into `ifs`, both single-use, to return to ten. The penalty math is unchanged.

sonarqubecloud · 2026-07-21T05:19:49Z

Quality Gate failed

Failed conditions
9.6% Duplication on New Code (required ≤ 3%)

See analysis details on SonarQube Cloud

0crat · 2026-07-21T13:34:43Z

@maxonfjvipon Thanks for the review! You've earned +24 points for this: +12 as a basis; +16 for the 11826 hits-of-code that you reviewed; -4 to not go over the cap. Your running score is +816; don't forget to check your Zerocracy account too).

0crat · 2026-07-21T16:05:50Z

@yegor256 Thanks for the contribution! You've earned +4 points for this: +16 as a basis; -4 for too many hits-of-code (11826 >= 100); -8 for way too many hits-of-code (11826 >= 400); -8 for the lack of code review; +8 to give you at least something. Please, keep them coming. Your running score is +2588; don't forget to check your Zerocracy account too).

Copilot AI review requested due to automatic review settings July 16, 2026 14:03

Copilot AI reviewed Jul 16, 2026

github-actions Bot added the core label Jul 16, 2026

yegor256 added 16 commits July 16, 2026 17:48

Merge remote-tracking branch 'origin/master' into 5577

6044f0e

Merge remote-tracking branch 'origin/master' into 5577

ce97098

Merge remote-tracking branch 'origin/master' into 5577

40d2eea

# Conflicts: # eo-printer/src/test/resources/org/eolang/printer/print-packs/yaml/times.yaml

Merge remote-tracking branch 'origin/master' into 5577

99e2e41

Merge remote-tracking branch 'origin/master' into 5577

16c44dd

style(#5577): re-format runtime sources with fixed printer

9f5ab09

Regenerate the canonical .eo sources now that the printer bugs (#5581, #5583, #5584, #5589, #5591) are fixed on master. The re-format is idempotent and the full eo-runtime suite passes (1551 tests, 0 failures, 0 errors).

style(#5577): re-format runtime sources at INDENT=3

5cc806c

Regenerate the canonical .eo sources under the lowered indentation penalty. The re-format is idempotent and the full eo-runtime suite passes (1551 tests, 0 failures, 0 errors).

Merge remote-tracking branch 'origin/master' into 5577

60a303a

Merge remote-tracking branch 'origin/master' into 5577

94dbe35

Merge remote-tracking branch 'origin/master' into 5577

33ef135

yegor256 requested a review from maxonfjvipon July 17, 2026 07:34

maxonfjvipon reviewed Jul 17, 2026

View reviewed changes

yegor256 added 6 commits July 17, 2026 16:56

Merge remote-tracking branch 'origin/master' into 5577

f9b736b

Merge remote-tracking branch 'origin/master' into 5577

ac1255b

yegor256 and others added 10 commits July 20, 2026 12:11

Merge pull request #5654 from objectionary/claude/eo-issue-5653-fuvm69

3bee2dd

#5653: scale the per-space penalty super-linearly by application length

yegor256 mentioned this pull request Jul 20, 2026

eo-printer fails to collapse --> marker to shorthand like ++> due to omission in Penalty.binding() #5657

Closed

claude and others added 6 commits July 20, 2026 11:12

Merge pull request #5658 from objectionary/claude/issue-5657-fix-ztpbuo

6409919

fix(#5657): treat throwing-test marker `-->` as a name binding

Merge remote-tracking branch 'origin/master' into 5577

9bd4270

Merge remote-tracking branch 'origin/5577' into 5577

cd1cd5c

maxonfjvipon suggested changes Jul 20, 2026

View reviewed changes

yegor256 added 6 commits July 20, 2026 18:58

Merge remote-tracking branch 'origin/master' into 5577

1e9788f

yegor256 merged commit db4273a into master Jul 21, 2026
24 of 27 checks passed

yegor256 deleted the 5577 branch July 21, 2026 05:27

Uh oh!

Conversation

yegor256 commented Jul 16, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Jul 16, 2026

⚠️ Benchmark Comparison Unavailable

Uh oh!

yegor256 commented Jul 17, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yegor256 commented Jul 20, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sonarqubecloud Bot commented Jul 21, 2026

Quality Gate failed

Uh oh!

Uh oh!

0crat commented Jul 21, 2026

Uh oh!

0crat commented Jul 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants