Support reading ScriptLoader source from stdin

[frostney]

Summary

• read ScriptLoader source from stdin when no path is provided or when is used explicitly
• support stdin across interpreted execution, bytecode execution, and bytecode emit flows
• document the new CLI usage and cover stdin input parsing with a native regression test

Testing

• ./build.pas loader tests
• ./build/Goccia.ScriptLoader.Input.Test
• printf 'const x = 2 + 2;\nx;\n' | ./build/ScriptLoader
• printf 'const x = 2 + 2;\nx;\n' | ./build/ScriptLoader -
• printf 'const x = 2 + 2;\nx;\n' | ./build/ScriptLoader --mode=bytecode
• printf 'const x = 2 + 2;\nx;\n' | ./build/ScriptLoader --emit --output=/tmp/goccia-stdin-test.gbc

Closes #129

Summary by CodeRabbit

• New Features

  • ScriptLoader and BenchmarkRunner accept stdin (pipe) as input; "-" or no path reads from stdin. CLI adds --help, treats unknown options as errors, and requires explicit --output when emitting bytecode from stdin.

• Documentation

  • README and docs updated with stdin examples for interpreted, bytecode and benchmark usage.

• Tests / CI

  • Added unit tests for stdin handling and CI smoke tests covering stdin for interpreter and bytecode paths.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 1c88eb36-c52c-44fc-b58c-95dbde680129

📥 Commits

Reviewing files that changed from the base of the PR and between ddf87c4 and b250a78.

📒 Files selected for processing (1)

• ScriptLoader.dpr

🚧 Files skipped from review as they are similar to previous changes (1)

• ScriptLoader.dpr

---

📝 Walkthrough

Walkthrough

Added stdin support across tools: new input unit and tests; ScriptLoader and BenchmarkRunner accept and process source from stdin (or -/no-path) via in-memory parse/compile flows; CLI option handling updated; docs and CI updated with stdin examples and smoke tests.

Changes

Cohort / File(s)	Summary
Stdin input unit & tests units/Goccia.ScriptLoader.Input.pas, units/Goccia.ScriptLoader.Input.Test.pas	New unit defines STDIN_FILE_NAME, STDIN_PATH_MARKER, IsStdinPath() and ReadSourceFromText(); tests cover line preservation, empty input, and stdin marker detection.
ScriptLoader program ScriptLoader.dpr	Refactored to operate on in-memory TStringList sources (ParseSource, CompileSource, RunSource), added RunScriptFromStdin, treat -/no-path as stdin, tightened --emit/--output error handling, and added --help/unknown-option handling.
BenchmarkRunner program BenchmarkRunner.dpr	Added Goccia.ScriptLoader.Input use; moved TReportSpec to top-level; added Collect*FromSource helpers for interpreted/bytecode flows and RunBenchmarksFromStdin; CLI now runs benchmarks from stdin when appropriate and improved option/error handling.
Documentation README.md, docs/build-system.md, docs/embedding.md, docs/benchmarks.md, AGENTS.md	Added examples and wording showing ScriptLoader/BenchmarkRunner accept files or stdin; updated embedding/build/benchmark docs and examples to demonstrate piping source via stdin.
CI / Workflows .github/workflows/ci.yml, .github/workflows/pr.yml	Added example job/steps that run stdin smoke tests: pipe JS into ./build/ScriptLoader (interpreted and --mode=bytecode) and into ./build/BenchmarkRunner to produce JSON outputs and assert benchmark presence.

Sequence Diagram(s)

sequenceDiagram
    participant User as User/CLI
    participant SL as ScriptLoader / BenchmarkRunner
    participant Input as Input Unit\n(IsStdinPath / ReadSourceFromText)
    participant Parser as Parser/Compiler
    participant Engine as Engine\n(Interpreter / Bytecode backend)
    participant Output as Output\n(stdout / file)

    User->>SL: Invoke with no args or `-` (stdin)
    SL->>Input: IsStdinPath(path)
    Input-->>SL: true
    SL->>Input: ReadSourceFromText(stdin)
    Input-->>SL: TStringList (source)
    SL->>Parser: ParseSource(ASource, AFileName)
    Parser-->>SL: AST / parse result
    alt mode = bytecode
        SL->>Parser: CompileSource(ASource, AFileName)
        Parser-->>Engine: Bytecode module
        Engine-->>Output: Emit bytecode / Run module (per flags)
    else mode = interpreted
        SL->>Engine: RunInterpreted(ASource, AFileName)
        Engine-->>Output: Execution result to stdout
    end

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• First draft of WAM output through Souffle VM #53 — touches ScriptLoader CLI and stdin/--emit/--output behaviors; strongly related to the in-memory/stdin refactor.
• Add bytecode and vm #38 — modifies interpreted vs. bytecode collection and execution paths; intersects with new Collect*FromSource logic.
• Potential Windows BenchmarkRunner fixes #62 — overlaps BenchmarkRunner collection/control-flow refactors and result propagation relevant to the new stdin collection functions.

Poem

🐇 I munch the bytes that piped on through,

From stdin's stream to engine new,
No temp file burrow, no dusty crate,
A hop, a parse, and code runs straight. 🥕

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Out of Scope Changes check	❓ Inconclusive	The pull request introduces changes to BenchmarkRunner.dpr (stdin support, new test procedures) and documentation that extend beyond the stated objective of 'Support reading ScriptLoader source from stdin' in the title. While aligned with issue #129 scope, BenchmarkRunner modifications broaden the scope beyond ScriptLoader alone.	Clarify whether BenchmarkRunner stdin support was an explicit requirement of issue #129 or if it represents scope expansion. If out of scope, consider separating into a follow-up PR for consistency with the primary objective.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The pull request title clearly and accurately summarizes the main change: adding stdin support to ScriptLoader. It is concise, specific, and aligns with the primary objective stated in issue #129.
Linked Issues check	✅ Passed	The pull request successfully implements all primary coding objectives from issue #129: stdin reading when no path/'-' provided [#129], virtual '' filename for error reporting [#129], flag preservation [#129], TStringList collection approach [#129], and --emit support with stdin [#129]. Documentation and tests comprehensively cover the implementation.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feature/scriptloader-stdin

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

units/Goccia.ScriptLoader.Input.Test.pas (1)

39-43: Add a fixture without a terminating newline.

The new CLI examples use printf "const x = 2 + 2; x;", which does not append a trailing newline. This setup writes the source with TStringList.SaveToFile, so it only exercises a newline-terminated last line and won't catch EOF-on-last-line regressions.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@units/Goccia.ScriptLoader.Input.Test.pas` around lines 39 - 43, The current
test uses TStringList.Create and SaveToFile (with TempFileName) which always
writes a trailing newline on the last line; replace that SaveToFile usage so the
fixture is written without a terminating newline — e.g. produce the file by
writing the two lines joined with sLineBreak but not ending with sLineBreak
using a raw writer (TFile.WriteAllText or similar) to TempFileName instead of
SaveToFile; update the block around TStringList.Create / Add / SaveToFile to
write "const x = 2 + 2;" + sLineBreak + "x;" directly to TempFileName so EOF has
no trailing newline.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@ScriptLoader.dpr`:
- Around line 409-416: When argument parsing leaves Paths.Count = 0 we must not
immediately fall back to RunScriptFromStdin; instead detect and reject
unknown/unsupported flags first. Update the argument-parsing flow so that after
parsing you track whether any unrecognized options were seen (or whether
exclusive flags like --help/--version were passed) and if so print a clear
error/help message and exit non‑zero; only call RunScriptFromStdin when no
unknown flags were present and no explicit paths were provided. Modify the
branch around Paths, RunScriptFromStdin and RunScripts to perform this
validation before defaulting to stdin.

---

Nitpick comments:
In `@units/Goccia.ScriptLoader.Input.Test.pas`:
- Around line 39-43: The current test uses TStringList.Create and SaveToFile
(with TempFileName) which always writes a trailing newline on the last line;
replace that SaveToFile usage so the fixture is written without a terminating
newline — e.g. produce the file by writing the two lines joined with sLineBreak
but not ending with sLineBreak using a raw writer (TFile.WriteAllText or
similar) to TempFileName instead of SaveToFile; update the block around
TStringList.Create / Add / SaveToFile to write "const x = 2 + 2;" + sLineBreak +
"x;" directly to TempFileName so EOF has no trailing newline.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: a4be2166-45cd-4e85-84c0-c5dc8a8428a6

📥 Commits

Reviewing files that changed from the base of the PR and between cf23908 and f285c72.

📒 Files selected for processing (6)

• README.md
• ScriptLoader.dpr
• docs/build-system.md
• docs/embedding.md
• units/Goccia.ScriptLoader.Input.Test.pas
• units/Goccia.ScriptLoader.Input.pas

[github-actions]

Benchmark Results

263 benchmarks

Interpreted: 🟢 47 improved · 216 unchanged · avg +4.5%
Bytecode: 🟢 14 improved · 🔴 10 regressed · 239 unchanged · avg -0.8%

arraybuffer.js — Interp: 14 unch. · avg +0.5% · Bytecode: 14 unch. · avg -0.9%

Benchmark	Interpreted	Δ	Bytecode	Δ
create ArrayBuffer(0)	460,941 → 475,967	+3.3%	495,931 → 488,941	-1.4%
create ArrayBuffer(64)	438,676 → 454,265	+3.6%	468,702 → 469,742	+0.2%
create ArrayBuffer(1024)	340,801 → 355,100	+4.2%	364,246 → 366,281	+0.6%
create ArrayBuffer(8192)	153,161 → 156,057	+1.9%	158,092 → 161,358	+2.1%
slice full buffer (64 bytes)	510,297 → 506,183	-0.8%	613,696 → 599,604	-2.3%
slice half buffer (512 of 1024 bytes)	443,995 → 428,211	-3.6%	514,002 → 516,243	+0.4%
slice with negative indices	420,545 → 418,271	-0.5%	556,195 → 553,796	-0.4%
slice empty range	492,633 → 474,396	-3.7%	586,081 → 590,685	+0.8%
byteLength access	1,373,383 → 1,395,704	+1.6%	1,617,053 → 1,553,218	-3.9%
Symbol.toStringTag access	1,058,950 → 1,077,335	+1.7%	1,086,785 → 1,030,007	-5.2%
ArrayBuffer.isView	775,339 → 755,947	-2.5%	891,671 → 868,656	-2.6%
clone ArrayBuffer(64)	419,168 → 420,389	+0.3%	463,484 → 464,317	+0.2%
clone ArrayBuffer(1024)	327,191 → 327,930	+0.2%	358,197 → 355,824	-0.7%
clone ArrayBuffer inside object	262,347 → 265,784	+1.3%	286,087 → 283,558	-0.9%

arrays.js — Interp: 🟢 1, 18 unch. · avg +2.1% · Bytecode: 🟢 3, 🔴 1, 15 unch. · avg +2.1%

Benchmark	Interpreted	Δ	Bytecode	Δ
Array.from length 100	13,491 → 13,401	-0.7%	14,184 → 14,376	+1.3%
Array.from 10 elements	214,985 → 223,490	+4.0%	241,622 → 233,878	-3.2%
Array.of 10 elements	275,483 → 282,343	+2.5%	302,710 → 298,373	-1.4%
spread into new array	322,851 → 331,540	+2.7%	196,204 → 180,473	🔴 -8.0%
map over 50 elements	24,635 → 24,865	+0.9%	27,151 → 26,855	-1.1%
filter over 50 elements	22,913 → 22,872	-0.2%	23,482 → 23,722	+1.0%
reduce sum 50 elements	26,112 → 26,677	+2.2%	16,056 → 19,692	🟢 +22.6%
forEach over 50 elements	23,121 → 23,389	+1.2%	27,528 → 28,574	+3.8%
find in 50 elements	33,145 → 32,534	-1.8%	30,392 → 30,573	+0.6%
sort 20 elements	12,449 → 12,431	-0.1%	13,457 → 13,685	+1.7%
flat nested array	109,704 → 117,819	🟢 +7.4%	127,573 → 120,374	-5.6%
flatMap	70,462 → 73,577	+4.4%	80,098 → 76,731	-4.2%
map inside map (5x5)	19,687 → 20,762	+5.5%	22,208 → 21,729	-2.2%
filter inside map (5x10)	15,317 → 15,824	+3.3%	16,265 → 16,220	-0.3%
reduce inside map (5x10)	18,791 → 19,265	+2.5%	13,585 → 16,294	🟢 +19.9%
forEach inside forEach (5x10)	16,662 → 16,970	+1.9%	14,260 → 17,243	🟢 +20.9%
find inside some (10x10)	13,505 → 13,385	-0.9%	12,942 → 12,949	+0.1%
map+filter chain nested (5x20)	5,102 → 5,189	+1.7%	5,436 → 5,411	-0.5%
reduce flatten (10x5)	37,206 → 38,416	+3.3%	17,488 → 16,657	-4.8%

async-await.js — Interp: 6 unch. · avg +4.6% · Bytecode: 6 unch. · avg -4.2%

Benchmark	Interpreted	Δ	Bytecode	Δ
single await	372,742 → 390,102	+4.7%	407,640 → 392,562	-3.7%
multiple awaits	167,530 → 174,545	+4.2%	176,902 → 172,755	-2.3%
await non-Promise value	854,598 → 886,010	+3.7%	1,102,890 → 1,035,024	-6.2%
await with try/catch	363,824 → 380,059	+4.5%	400,395 → 387,848	-3.1%
await Promise.all	49,829 → 52,279	+4.9%	50,490 → 48,162	-4.6%
nested async function call	189,559 → 200,008	+5.5%	231,699 → 219,597	-5.2%

classes.js — Interp: 🟢 1, 30 unch. · avg +4.0% · Bytecode: 🟢 1, 🔴 1, 29 unch. · avg -2.1%

Benchmark	Interpreted	Δ	Bytecode	Δ
simple class new	139,342 → 140,757	+1.0%	206,440 → 199,117	-3.5%
class with defaults	109,652 → 114,945	+4.8%	145,737 → 139,213	-4.5%
50 instances via Array.from	5,562 → 5,701	+2.5%	7,938 → 7,773	-2.1%
instance method call	70,321 → 71,588	+1.8%	94,011 → 91,812	-2.3%
static method call	113,631 → 116,434	+2.5%	186,232 → 178,822	-4.0%
single-level inheritance	55,776 → 56,369	+1.1%	74,674 → 72,720	-2.6%
two-level inheritance	48,062 → 48,420	+0.7%	60,688 → 58,797	-3.1%
private field access	71,584 → 73,680	+2.9%	94,112 → 92,362	-1.9%
private methods	78,592 → 80,310	+2.2%	101,038 → 97,823	-3.2%
getter/setter access	74,869 → 76,690	+2.4%	102,495 → 103,416	+0.9%
class decorator (identity)	92,129 → 94,990	+3.1%	100,445 → 98,277	-2.2%
class decorator (wrapping)	54,286 → 55,748	+2.7%	57,766 → 57,087	-1.2%
identity method decorator	64,211 → 66,690	+3.9%	85,710 → 87,876	+2.5%
wrapping method decorator	51,999 → 53,476	+2.8%	59,926 → 58,129	-3.0%
stacked method decorators (x3)	34,713 → 36,279	+4.5%	41,715 → 39,827	-4.5%
identity field decorator	70,189 → 73,276	+4.4%	73,754 → 70,620	-4.2%
field initializer decorator	58,673 → 60,803	+3.6%	62,262 → 60,270	-3.2%
getter decorator (identity)	63,775 → 67,914	+6.5%	75,653 → 70,741	-6.5%
setter decorator (identity)	53,636 → 56,264	+4.9%	59,111 → 58,355	-1.3%
static method decorator	69,756 → 72,640	+4.1%	81,194 → 77,465	-4.6%
static field decorator	80,826 → 83,898	+3.8%	87,932 → 83,076	-5.5%
private method decorator	52,600 → 54,832	+4.2%	68,372 → 69,393	+1.5%
private field decorator	58,469 → 60,592	+3.6%	62,322 → 60,595	-2.8%
plain auto-accessor (no decorator)	101,091 → 105,312	+4.2%	106,551 → 95,048	🔴 -10.8%
auto-accessor with decorator	53,256 → 56,517	+6.1%	54,970 → 61,202	🟢 +11.3%
decorator writing metadata	43,531 → 46,159	+6.0%	50,266 → 47,028	-6.4%
static getter read	123,808 → 133,662	🟢 +8.0%	169,328 → 172,542	+1.9%
static getter/setter pair	95,900 → 102,343	+6.7%	123,501 → 118,274	-4.2%
inherited static getter	75,569 → 80,096	+6.0%	95,570 → 98,569	+3.1%
inherited static setter	80,798 → 85,836	+6.2%	104,821 → 102,772	-2.0%
inherited static getter with this binding	69,205 → 72,691	+5.0%	90,039 → 92,027	+2.2%

closures.js — Interp: 🟢 1, 10 unch. · avg +4.4% · Bytecode: 🔴 1, 10 unch. · avg -0.9%

Benchmark	Interpreted	Δ	Bytecode	Δ
closure over single variable	133,548 → 141,911	+6.3%	277,191 → 277,976	+0.3%
closure over multiple variables	116,283 → 126,344	🟢 +8.7%	194,151 → 197,225	+1.6%
nested closures	124,452 → 131,501	+5.7%	229,882 → 225,698	-1.8%
function as argument	91,538 → 94,501	+3.2%	254,280 → 254,203	-0.0%
function returning function	116,989 → 121,844	+4.2%	282,700 → 285,148	+0.9%
compose two functions	70,113 → 72,068	+2.8%	151,223 → 151,782	+0.4%
fn.call	147,731 → 155,720	+5.4%	231,913 → 230,753	-0.5%
fn.apply	106,906 → 110,767	+3.6%	159,550 → 146,420	🔴 -8.2%
fn.bind	141,580 → 146,915	+3.8%	285,902 → 282,343	-1.2%
recursive sum to 50	11,965 → 12,233	+2.2%	23,993 → 24,215	+0.9%
recursive tree traversal	20,023 → 20,612	+2.9%	36,071 → 35,267	-2.2%

collections.js — Interp: 🟢 2, 10 unch. · avg +4.9% · Bytecode: 🟢 3, 9 unch. · avg +2.8%

Benchmark	Interpreted	Δ	Bytecode	Δ
add 50 elements	6,988 → 7,254	+3.8%	7,221 → 7,283	+0.9%
has lookup (50 elements)	89,257 → 91,190	+2.2%	99,831 → 98,791	-1.0%
delete elements	46,975 → 48,192	+2.6%	49,035 → 49,597	+1.1%
forEach iteration	14,948 → 15,849	+6.0%	12,681 → 15,033	🟢 +18.5%
spread to array	27,496 → 29,605	🟢 +7.7%	216,816 → 211,516	-2.4%
deduplicate array	36,344 → 38,488	+5.9%	70,451 → 71,090	+0.9%
set 50 entries	5,084 → 5,551	🟢 +9.2%	5,299 → 5,672	🟢 +7.0%
get lookup (50 entries)	85,887 → 89,481	+4.2%	93,754 → 91,886	-2.0%
has check	127,840 → 132,932	+4.0%	144,729 → 141,131	-2.5%
delete entries	45,662 → 47,396	+3.8%	47,428 → 47,747	+0.7%
forEach iteration	15,114 → 15,695	+3.8%	12,881 → 15,067	🟢 +17.0%
keys/values/entries	7,417 → 7,834	+5.6%	27,566 → 26,213	-4.9%

destructuring.js — Interp: 🟢 5, 17 unch. · avg +5.2% · Bytecode: 🟢 3, 🔴 1, 18 unch. · avg -0.9%

Benchmark	Interpreted	Δ	Bytecode	Δ
simple array destructuring	366,014 → 384,929	+5.2%	266,798 → 259,231	-2.8%
with rest element	239,209 → 246,627	+3.1%	204,028 → 199,078	-2.4%
with defaults	373,070 → 399,520	🟢 +7.1%	309,488 → 295,614	-4.5%
skip elements	385,015 → 407,377	+5.8%	281,898 → 271,019	-3.9%
nested array destructuring	165,270 → 172,351	+4.3%	90,160 → 84,964	-5.8%
swap variables	488,633 → 504,462	+3.2%	328,616 → 319,213	-2.9%
simple object destructuring	297,504 → 310,028	+4.2%	401,228 → 397,162	-1.0%
with defaults	329,763 → 351,242	+6.5%	563,623 → 539,544	-4.3%
with renaming	311,250 → 335,789	🟢 +7.9%	432,410 → 432,648	+0.1%
nested object destructuring	144,447 → 155,354	🟢 +7.6%	200,714 → 191,248	-4.7%
rest properties	177,696 → 191,317	🟢 +7.7%	182,029 → 182,253	+0.1%
object parameter	90,888 → 94,914	+4.4%	154,204 → 150,793	-2.2%
array parameter	116,529 → 122,984	+5.5%	109,915 → 105,786	-3.8%
mixed destructuring in map	32,821 → 33,368	+1.7%	29,934 → 35,134	🟢 +17.4%
forEach with array destructuring	60,737 → 63,653	+4.8%	41,506 → 40,407	-2.6%
map with array destructuring	59,916 → 64,429	🟢 +7.5%	41,247 → 40,107	-2.8%
filter with array destructuring	62,665 → 65,988	+5.3%	48,238 → 43,835	🔴 -9.1%
reduce with array destructuring	69,029 → 73,781	+6.9%	44,474 → 44,603	+0.3%
map with object destructuring	71,503 → 73,722	+3.1%	78,320 → 87,928	🟢 +12.3%
map with nested destructuring	60,065 → 61,535	+2.4%	87,338 → 83,333	-4.6%
map with rest in destructuring	40,139 → 42,515	+5.9%	23,025 → 22,938	-0.4%
map with defaults in destructuring	55,243 → 57,192	+3.5%	59,440 → 63,856	🟢 +7.4%

fibonacci.js — Interp: 8 unch. · avg +2.9% · Bytecode: 🔴 1, 7 unch. · avg -2.0%

Benchmark	Interpreted	Δ	Bytecode	Δ
recursive fib(15)	315 → 323	+2.6%	846 → 849	+0.4%
recursive fib(20)	29 → 29	-0.1%	76 → 76	+0.1%
recursive fib(15) typed	320 → 328	+2.3%	891 → 876	-1.7%
recursive fib(20) typed	29 → 29	+0.6%	80 → 80	+0.0%
iterative fib(20) via reduce	11,615 → 12,284	+5.8%	16,557 → 17,071	+3.1%
iterator fib(20)	9,091 → 9,511	+4.6%	18,482 → 17,484	-5.4%
iterator fib(20) via Iterator.from + take	14,714 → 15,278	+3.8%	20,763 → 19,781	-4.7%
iterator fib(20) last value via reduce	11,125 → 11,535	+3.7%	14,678 → 13,524	🔴 -7.9%

for-of.js — Interp: 🟢 7 · avg +8.7% · Bytecode: 7 unch. · avg -0.6%

Benchmark	Interpreted	Δ	Bytecode	Δ
for...of with 10-element array	43,456 → 47,602	🟢 +9.5%	165,570 → 165,616	+0.0%
for...of with 100-element array	5,000 → 5,430	🟢 +8.6%	17,595 → 17,921	+1.8%
for...of with string (10 chars)	32,204 → 35,055	🟢 +8.9%	87,390 → 87,082	-0.4%
for...of with Set (10 elements)	43,265 → 47,204	🟢 +9.1%	150,588 → 157,037	+4.3%
for...of with Map entries (10 entries)	27,939 → 30,329	🟢 +8.6%	28,324 → 27,198	-4.0%
for...of with destructuring	37,468 → 40,507	🟢 +8.1%	36,251 → 34,749	-4.1%
for-await-of with sync array	41,676 → 45,170	🟢 +8.4%	127,294 → 125,190	-1.7%

helpers/bench-module.js — Interp: 0 · Bytecode: 0

Benchmark	Interpreted	Δ	Bytecode	Δ

iterators.js — Interp: 🟢 5, 15 unch. · avg +5.7% · Bytecode: 🔴 1, 19 unch. · avg -2.0%

Benchmark	Interpreted	Δ	Bytecode	Δ
Iterator.from({next}).toArray() — 20 elements	14,588 → 15,520	+6.4%	20,343 → 18,681	🔴 -8.2%
Iterator.from({next}).toArray() — 50 elements	6,383 → 6,768	+6.0%	8,889 → 8,551	-3.8%
spread pre-wrapped iterator — 20 elements	10,757 → 11,897	🟢 +10.6%	19,662 → 18,314	-6.9%
Iterator.from({next}).forEach — 50 elements	4,386 → 4,614	+5.2%	5,085 → 5,439	+6.9%
Iterator.from({next}).reduce — 50 elements	4,457 → 4,695	+5.4%	5,129 → 4,997	-2.6%
wrap array iterator	153,764 → 167,179	🟢 +8.7%	177,396 → 176,266	-0.6%
wrap plain {next()} object	10,201 → 11,006	🟢 +7.9%	14,201 → 13,726	-3.3%
map + toArray (50 elements)	4,409 → 4,683	+6.2%	5,895 → 5,923	+0.5%
filter + toArray (50 elements)	4,288 → 4,590	🟢 +7.0%	5,778 → 5,589	-3.3%
take(10) + toArray (50 element source)	26,452 → 28,345	🟢 +7.2%	35,683 → 33,445	-6.3%
drop(40) + toArray (50 element source)	6,392 → 6,735	+5.4%	8,925 → 8,677	-2.8%
chained map + filter + take (100 element source)	7,902 → 8,226	+4.1%	10,301 → 9,764	-5.2%
some + every (50 elements)	2,504 → 2,661	+6.3%	3,438 → 3,419	-0.6%
find (50 elements)	5,424 → 5,758	+6.2%	7,021 → 6,780	-3.4%
array.values().map().filter().toArray()	7,846 → 8,000	+2.0%	9,519 → 9,548	+0.3%
array.values().take(5).toArray()	203,427 → 213,983	+5.2%	233,122 → 230,087	-1.3%
array.values().drop(45).toArray()	187,752 → 195,937	+4.4%	219,365 → 211,589	-3.5%
map.entries() chained helpers	9,660 → 9,850	+2.0%	5,847 → 5,800	-0.8%
set.values() chained helpers	16,248 → 16,897	+4.0%	19,532 → 19,689	+0.8%
string iterator map + toArray	12,971 → 13,533	+4.3%	13,823 → 14,505	+4.9%

json.js — Interp: 🟢 1, 19 unch. · avg +4.4% · Bytecode: 20 unch. · avg +0.4%

Benchmark	Interpreted	Δ	Bytecode	Δ
parse simple object	166,039 → 176,068	+6.0%	174,262 → 179,418	+3.0%
parse nested object	102,140 → 108,417	+6.1%	108,870 → 108,501	-0.3%
parse array of objects	58,901 → 63,977	🟢 +8.6%	61,076 → 64,264	+5.2%
parse large flat object	66,386 → 68,849	+3.7%	69,452 → 70,819	+2.0%
parse mixed types	75,794 → 79,731	+5.2%	78,416 → 79,056	+0.8%
stringify simple object	191,242 → 199,280	+4.2%	215,730 → 216,679	+0.4%
stringify nested object	105,000 → 111,663	+6.3%	114,781 → 112,861	-1.7%
stringify array of objects	60,285 → 62,385	+3.5%	62,744 → 63,434	+1.1%
stringify mixed types	85,300 → 89,515	+4.9%	88,832 → 88,481	-0.4%
reviver doubles numbers	47,713 → 49,007	+2.7%	53,752 → 52,275	-2.7%
reviver filters properties	40,418 → 41,165	+1.8%	43,220 → 42,465	-1.7%
reviver on nested object	51,258 → 51,904	+1.3%	55,197 → 54,415	-1.4%
reviver on array	31,615 → 33,049	+4.5%	31,171 → 32,570	+4.5%
replacer function doubles numbers	46,093 → 47,766	+3.6%	52,009 → 52,481	+0.9%
replacer function excludes properties	58,701 → 61,512	+4.8%	63,657 → 62,339	-2.1%
array replacer (allowlist)	108,686 → 114,895	+5.7%	116,253 → 118,789	+2.2%
stringify with 2-space indent	93,325 → 97,328	+4.3%	103,446 → 101,940	-1.5%
stringify with tab indent	95,465 → 97,811	+2.5%	102,399 → 102,870	+0.5%
parse then stringify	54,591 → 57,316	+5.0%	59,858 → 59,803	-0.1%
stringify then parse	33,645 → 34,895	+3.7%	36,672 → 36,395	-0.8%

jsx.jsx — Interp: 🟢 2, 19 unch. · avg +5.1% · Bytecode: 🔴 4, 17 unch. · avg -5.3%

Benchmark	Interpreted	Δ	Bytecode	Δ
simple element	205,694 → 212,397	+3.3%	343,642 → 317,989	🔴 -7.5%
self-closing element	213,942 → 222,559	+4.0%	353,831 → 339,248	-4.1%
element with string attribute	168,319 → 178,109	+5.8%	264,417 → 241,209	🔴 -8.8%
element with multiple attributes	150,805 → 157,765	+4.6%	203,139 → 196,718	-3.2%
element with expression attribute	156,864 → 168,069	🟢 +7.1%	263,450 → 241,037	🔴 -8.5%
text child	205,289 → 213,487	+4.0%	325,773 → 312,323	-4.1%
expression child	198,789 → 210,069	+5.7%	342,691 → 315,501	🔴 -7.9%
mixed text and expression	187,434 → 198,316	+5.8%	293,787 → 284,242	-3.2%
nested elements (3 levels)	76,802 → 79,760	+3.9%	124,856 → 116,461	-6.7%
sibling children	57,425 → 59,907	+4.3%	87,499 → 83,676	-4.4%
component element	142,098 → 151,063	+6.3%	232,923 → 216,624	-7.0%
component with children	88,778 → 93,062	+4.8%	138,516 → 130,387	-5.9%
dotted component	119,804 → 127,254	+6.2%	177,898 → 171,554	-3.6%
empty fragment	214,444 → 220,751	+2.9%	370,780 → 348,630	-6.0%
fragment with children	56,993 → 59,578	+4.5%	91,710 → 87,846	-4.2%
spread attributes	104,647 → 112,205	🟢 +7.2%	132,465 → 126,826	-4.3%
spread with overrides	91,442 → 95,572	+4.5%	118,247 → 114,929	-2.8%
shorthand props	152,893 → 161,831	+5.8%	237,292 → 226,108	-4.7%
nav bar structure	26,189 → 27,998	+6.9%	40,637 → 38,519	-5.2%
card component tree	30,935 → 32,197	+4.1%	45,123 → 43,036	-4.6%
10 list items via Array.from	13,839 → 14,656	+5.9%	19,357 → 18,506	-4.4%

modules.js — Interp: 9 unch. · avg +0.7% · Bytecode: 9 unch. · avg +2.1%

Benchmark	Interpreted	Δ	Bytecode	Δ
call imported function	464,381 → 479,176	+3.2%	633,078 → 669,863	+5.8%
call two imported functions	260,650 → 267,281	+2.5%	360,261 → 367,635	+2.0%
read imported constant	1,658,124 → 1,677,642	+1.2%	3,649,685 → 3,546,700	-2.8%
read imported string	1,660,082 → 1,660,298	+0.0%	3,621,450 → 3,621,405	-0.0%
read JSON string property	1,676,841 → 1,672,593	-0.3%	3,629,054 → 3,709,278	+2.2%
read JSON number property	1,648,678 → 1,627,482	-1.3%	3,616,919 → 3,660,199	+1.2%
read JSON boolean property	1,668,628 → 1,674,827	+0.4%	3,555,509 → 3,622,079	+1.9%
read JSON array property	1,657,885 → 1,666,110	+0.5%	3,546,051 → 3,718,797	+4.9%
read multiple JSON properties	947,909 → 944,678	-0.3%	3,194,278 → 3,319,671	+3.9%

numbers.js — Interp: 🟢 3, 8 unch. · avg +5.9% · Bytecode: 11 unch. · avg +0.5%

Benchmark	Interpreted	Δ	Bytecode	Δ
integer arithmetic	502,885 → 522,900	+4.0%	1,456,212 → 1,515,396	+4.1%
floating point arithmetic	579,536 → 620,998	🟢 +7.2%	945,920 → 969,925	+2.5%
number coercion	178,304 → 188,892	+5.9%	235,844 → 224,693	-4.7%
toFixed	99,458 → 103,998	+4.6%	110,863 → 111,370	+0.5%
toString	150,794 → 160,155	+6.2%	182,808 → 184,394	+0.9%
valueOf	221,634 → 240,281	🟢 +8.4%	260,666 → 265,122	+1.7%
toPrecision	141,238 → 148,857	+5.4%	164,957 → 163,361	-1.0%
Number.isNaN	315,858 → 334,046	+5.8%	340,023 → 341,793	+0.5%
Number.isFinite	309,679 → 329,845	+6.5%	324,616 → 323,816	-0.2%
Number.isInteger	319,143 → 342,846	🟢 +7.4%	345,748 → 345,745	-0.0%
Number.parseInt and parseFloat	248,754 → 256,833	+3.2%	262,008 → 264,070	+0.8%

objects.js — Interp: 🟢 2, 5 unch. · avg +5.5% · Bytecode: 7 unch. · avg -0.9%

Benchmark	Interpreted	Δ	Bytecode	Δ
create simple object	456,850 → 483,270	+5.8%	580,784 → 565,688	-2.6%
create nested object	218,686 → 236,428	🟢 +8.1%	249,296 → 250,547	+0.5%
create 50 objects via Array.from	8,803 → 9,275	+5.4%	9,816 → 9,783	-0.3%
property read	542,708 → 573,785	+5.7%	427,376 → 443,676	+3.8%
Object.keys	270,021 → 282,363	+4.6%	300,881 → 302,652	+0.6%
Object.entries	99,432 → 101,089	+1.7%	107,094 → 103,416	-3.4%
spread operator	169,210 → 181,418	🟢 +7.2%	208,794 → 198,969	-4.7%

promises.js — Interp: 🟢 6, 6 unch. · avg +7.0% · Bytecode: 12 unch. · avg -2.8%

Benchmark	Interpreted	Δ	Bytecode	Δ
Promise.resolve(value)	528,220 → 555,732	+5.2%	570,312 → 538,244	-5.6%
new Promise(resolve => resolve(value))	181,819 → 191,020	+5.1%	223,033 → 215,001	-3.6%
Promise.reject(reason)	531,483 → 566,057	+6.5%	559,418 → 529,143	-5.4%
resolve + then (1 handler)	169,489 → 180,239	+6.3%	197,428 → 192,764	-2.4%
resolve + then chain (3 deep)	66,284 → 71,997	🟢 +8.6%	75,956 → 76,458	+0.7%
resolve + then chain (10 deep)	21,338 → 23,192	🟢 +8.7%	23,444 → 24,122	+2.9%
reject + catch + then	96,104 → 101,306	+5.4%	111,509 → 108,440	-2.8%
resolve + finally + then	82,229 → 85,580	+4.1%	91,036 → 88,941	-2.3%
Promise.all (5 resolved)	29,931 → 32,300	🟢 +7.9%	33,157 → 31,404	-5.3%
Promise.race (5 resolved)	31,833 → 34,334	🟢 +7.9%	34,410 → 33,154	-3.7%
Promise.allSettled (5 mixed)	25,148 → 27,288	🟢 +8.5%	27,069 → 26,807	-1.0%
Promise.any (5 mixed)	29,745 → 32,801	🟢 +10.3%	32,791 → 31,010	-5.4%

strings.js — Interp: 🟢 2, 9 unch. · avg +4.9% · Bytecode: 11 unch. · avg -1.3%

Benchmark	Interpreted	Δ	Bytecode	Δ
string concatenation	424,231 → 423,545	-0.2%	312,204 → 318,782	+2.1%
template literal	620,903 → 692,675	🟢 +11.6%	652,069 → 670,298	+2.8%
string repeat	394,169 → 406,833	+3.2%	462,802 → 448,921	-3.0%
split and join	132,757 → 139,840	+5.3%	156,701 → 154,909	-1.1%
indexOf and includes	162,234 → 170,307	+5.0%	192,158 → 187,199	-2.6%
toUpperCase and toLowerCase	245,867 → 265,942	🟢 +8.2%	334,120 → 324,950	-2.7%
slice and substring	150,016 → 156,724	+4.5%	197,358 → 195,118	-1.1%
trim operations	185,684 → 194,404	+4.7%	243,704 → 237,495	-2.5%
replace and replaceAll	206,727 → 213,771	+3.4%	240,202 → 236,154	-1.7%
startsWith and endsWith	131,971 → 138,141	+4.7%	159,326 → 155,804	-2.2%
padStart and padEnd	189,141 → 196,048	+3.7%	230,666 → 226,524	-1.8%

typed-arrays.js — Interp: 🟢 9, 13 unch. · avg +5.2% · Bytecode: 🟢 4, 18 unch. · avg +2.4%

Benchmark	Interpreted	Δ	Bytecode	Δ
new Int32Array(0)	327,452 → 336,169	+2.7%	364,303 → 355,533	-2.4%
new Int32Array(100)	297,472 → 305,445	+2.7%	329,045 → 329,173	+0.0%
new Int32Array(1000)	177,496 → 186,932	+5.3%	196,587 → 196,467	-0.1%
new Float64Array(100)	265,195 → 287,131	🟢 +8.3%	305,726 → 303,286	-0.8%
Int32Array.from([...])	175,065 → 183,769	+5.0%	191,198 → 188,271	-1.5%
Int32Array.of(1, 2, 3, 4, 5)	292,592 → 314,180	🟢 +7.4%	355,822 → 347,423	-2.4%
sequential write 100 elements	3,343 → 3,614	🟢 +8.1%	15,106 → 16,856	🟢 +11.6%
sequential read 100 elements	3,388 → 3,716	🟢 +9.7%	10,517 → 11,180	+6.3%
Float64Array write 100 elements	3,108 → 3,364	🟢 +8.2%	10,821 → 11,700	🟢 +8.1%
fill(42)	43,916 → 45,725	+4.1%	44,603 → 45,949	+3.0%
slice()	196,584 → 203,671	+3.6%	220,649 → 219,050	-0.7%
map(x => x * 2)	7,553 → 8,087	🟢 +7.1%	8,645 → 8,732	+1.0%
filter(x => x > 50)	7,804 → 8,363	🟢 +7.2%	9,546 → 9,444	-1.1%
reduce (sum)	7,624 → 7,923	+3.9%	6,362 → 7,522	🟢 +18.2%
sort()	164,971 → 168,180	+1.9%	181,147 → 178,693	-1.4%
indexOf()	419,068 → 417,640	-0.3%	485,535 → 492,378	+1.4%
reverse()	319,863 → 326,313	+2.0%	356,545 → 356,508	-0.0%
create view over existing buffer	394,227 → 391,959	-0.6%	444,280 → 431,793	-2.8%
subarray()	408,053 → 415,035	+1.7%	517,670 → 521,207	+0.7%
set() from array	545,389 → 570,298	+4.6%	646,635 → 646,547	-0.0%
for-of loop	4,708 → 5,150	🟢 +9.4%	17,578 → 19,417	🟢 +10.5%
spread into array	15,730 → 17,530	🟢 +11.4%	95,027 → 100,646	+5.9%

_{Measured on ubuntu-latest x64. Changes within ±7% are considered insignificant.}

[github-actions]

Suite Timing

Suite	Metric	Interpreted	Bytecode
Tests	Total	3523	3523
Tests	Passed	3482 ✅	3523 ✅
Tests	Skipped	41	0
Tests	Execution	158.1ms	140.4ms
Tests	Engine	317.3ms	515.9ms
Benchmarks	Total	263	263
Benchmarks	Duration	7.27min	5.98min

_{Measured on ubuntu-latest x64.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

.github/workflows/ci.yml (1)

332-346: Please cover the remaining stdin paths in CI.

These smoke tests verify interpreted and bytecode execution, but the new Goccia.ScriptLoader.Input.Test binary is still only built—not run—and stdin --emit --output=... is the one new ScriptLoader flow still missing automated coverage.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In @.github/workflows/ci.yml around lines 332 - 346, CI lacks execution of the
new Goccia.ScriptLoader.Input.Test binary and the ScriptLoader stdin flow that
uses --emit --output; update the workflow to run the built
Goccia.ScriptLoader.Input.Test binary (e.g.,
./build/Goccia.ScriptLoader.Input.Test) to exercise its stdin path, and add a
smoke step that pipes a small script into ./build/ScriptLoader using the --emit
--output=... flag (pointing at a temp file), then assert the emitted output file
exists and contains expected content (or that subsequent bytecode run produces
"Result: 4"); reference Goccia.ScriptLoader.Input.Test and the ScriptLoader
--emit --output stdin flow when adding these checks.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@ScriptLoader.dpr`:
- Around line 282-289: RunScriptFromFile is unconditionally loading every
AFileName as text which corrupts .gbc bytecode; before creating/LoadFromFile,
detect the file extension of AFileName (case-insensitive) and if it's ".gbc"
call RunBytecodeFromFile(AFileName) and return; only for non-.gbc paths continue
to create the TStringList, LoadFromFile and call RunSource(Source, AFileName).
Ensure you perform the extension check early in RunScriptFromFile so bytecode
files never go through TStringList.LoadFromFile.

---

Nitpick comments:
In @.github/workflows/ci.yml:
- Around line 332-346: CI lacks execution of the new
Goccia.ScriptLoader.Input.Test binary and the ScriptLoader stdin flow that uses
--emit --output; update the workflow to run the built
Goccia.ScriptLoader.Input.Test binary (e.g.,
./build/Goccia.ScriptLoader.Input.Test) to exercise its stdin path, and add a
smoke step that pipes a small script into ./build/ScriptLoader using the --emit
--output=... flag (pointing at a temp file), then assert the emitted output file
exists and contains expected content (or that subsequent bytecode run produces
"Result: 4"); reference Goccia.ScriptLoader.Input.Test and the ScriptLoader
--emit --output stdin flow when adding these checks.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 842e4df1-2a90-4a20-9c24-af2a3d96fb3c

📥 Commits

Reviewing files that changed from the base of the PR and between f285c72 and ddf87c4.

📒 Files selected for processing (10)

• .github/workflows/ci.yml
• .github/workflows/pr.yml
• AGENTS.md
• BenchmarkRunner.dpr
• README.md
• ScriptLoader.dpr
• docs/benchmarks.md
• docs/build-system.md
• docs/embedding.md
• units/Goccia.ScriptLoader.Input.Test.pas

✅ Files skipped from review due to trivial changes (4)

• docs/benchmarks.md
• README.md
• .github/workflows/pr.yml
• docs/build-system.md

🚧 Files skipped from review as they are similar to previous changes (1)

• docs/embedding.md