SICPy: large amounts of untranslated JavaScript still sitting in <PYTHON> tags (mostly ch5.5 compiler)

[martin-henz]

Summary

While addressing Gemini's review on #1304 (which caught 17 error() calls inside still-untranslated JavaScript blocks), I scanned all of xml_py for the same underlying problem: content inside <PYTHON>/<PYTHONINLINE>/<PYTHON_RUN>/<PYTHON_TEST> tags that is still literal, untranslated JavaScript (function declarations, const/let, ternaries, semicolons, ===) rather than real Python.

This is much bigger than the 17 lines Gemini flagged — roughly 90 code blocks across 19 files, heavily concentrated in the chapter 5.5 compiler section. One file (xml_py/chapter5/section3/subsection2.xml) had its entire PYTHON_RUN block still in JS — all 13 of its error() calls were flagged.

There's already a FIXME comment in the source acknowledging this in at least one spot (xml_py/chapter5/section5/subsection2.xml:796: <!-- FIXME: This SPLIT.JAVASCRIPT contains a lot of SCHEME. Remove. -->), so this is a known, not-yet-finished conversion gap rather than a surprise.

Methodology (and caveats)

Regex scan requiring at least 2 distinct JS-only signals per code block (function name(, const x =, let x =, =>, ===, !==, trailing ;, ternary a ? b : c). This has some false positives from prose that happens to use the English word "function" or contain ?/: punctuation near each other (e.g. exercise text, subheadings) — I filtered out the ones I could identify manually, but a human skim is still worth it before treating the full list as ground truth. It should not have significant false negatives for genuine top-level function/const/let code, which is the bulk of what's below.

Two distinct shapes of the same problem:

1. Implementation code still written as JS (function bodies, register-machine ops, etc.) — needs real translation to Python (def, no semicolons, if/else instead of ternaries, etc.)
2. Demonstration input strings (e.g. parse_and_evaluate("... function factorial(n) {..."), compile(parse(function factorial...))) — these feed a JS-syntax program into an evaluator/compiler being demonstrated; for the Python edition these example programs should presumably also be Python syntax once translated.

Representative examples (genuine code, not prose)

xml_py/chapter5/section3/subsection2.xml:817   // TYPED POINTERS  const NUMBER_TYPE = "number"; ...
xml_py/chapter5/section5/subsection2.xml:1021  function compile_lambda_expression(exp, target, linkage) { const fun_entry = ...
xml_py/chapter5/section5/subsection3.xml:603   function compile_function_call(target, linkage) { const primitive_branch = ...
xml_py/chapter5/section4/subsection4.xml:1170  function fib(n) { return n < 2 ? n : fib(n - 1) + fib(n - 2); }
xml_py/chapter4/section1/subsection6.xml:360   function f(x) { return ((is_even, is_odd) => is_even(is_even, is_odd, x))...
xml_py/chapter3/section3/subsection1.xml:248   const z = pair(y, tail(x));

Rough distribution (block count per file, after removing obvious prose false-positives)

 2  xml_py/chapter2/section3/subsection2.xml
 3  xml_py/chapter3/section3/subsection1.xml
 2  xml_py/chapter4/section1/subsection2.xml
 4  xml_py/chapter4/section1/subsection4.xml
 2  xml_py/chapter4/section1/subsection6.xml
 2  xml_py/chapter4/section1/subsection7.xml
 7  xml_py/chapter4/section2/subsection2.xml
 1  xml_py/chapter4/section3/section3.xml
 8  xml_py/chapter4/section3/subsection3.xml
 3  xml_py/chapter5/section3/subsection1.xml
 1  xml_py/chapter5/section3/subsection2.xml   (only 1 block, but that block is the file's entire PYTHON_RUN driver-loop/simulator)
 5  xml_py/chapter5/section4/subsection2.xml
13  xml_py/chapter5/section4/subsection4.xml
 2  xml_py/chapter5/section5/subsection1.xml
13  xml_py/chapter5/section5/subsection2.xml
 6  xml_py/chapter5/section5/subsection3.xml
 7  xml_py/chapter5/section5/subsection4.xml
 4  xml_py/chapter5/section5/subsection5.xml
10  xml_py/chapter5/section5/subsection7.xml

~42 of the ~90 blocks are in chapter 5.5 alone (the compiler — subsections 1 through 7), which reads as essentially not yet translated to Python. Chapters 5.3/5.4 (register-machine simulator, explicit-control evaluator) and scattered spots in chapters 2–4 make up the rest.

Suggested next step

This is a content-translation task, not a mechanical fix — I haven't attempted it and don't think it's something to script the way the argument-order fix was. Given the size (most of chapter 5.5), it's probably worth scoping/prioritizing deliberately rather than doing it as a drive-by fix.