Convert JS strings to use template literals. NFC

[kleisauke]

Marked as draft as this seems to regress code size, so perhaps it should not be done everywhere.

[kleisauke]

Commit 55a68e8 was automatically done with the Biome linter:

$ find src -type f -name '*.js' -exec sed -i -E -e '/^ *#(if|ifdef|elif|else|endif|include|warning|error|preprocess)/d' -e 's/[\{<]{3} .* [\}>]{3}//g' {} +
$ git add src
$ git commit -m"[Temp] Remove pre-process syntax"
$ npx @biomejs/biome lint --only=lint/style/useTemplate --write --unsafe --skip-parse-errors ./src
$ git add src
$ git commit -m"Convert JS strings to use template literals. NFC"

(see: https://biomejs.dev/linter/rules/use-template/)

Though, that doesn't fix all files due to Emscripten's pre-process syntax, so I also ran:

$ ./emcc test/hello_world.c -sINCLUDE_FULL_LIBRARY
$ npx @biomejs/biome lint --only=lint/style/useTemplate a.out.js

And then manually applied the fixes reported by that (in commit 25c22b5).

FWIW, the lint/complexity/useArrowFunction and lint/nursery/useNullishCoalescing lint rules might also be useful as follow-ups.

[sbc100]

Oh nice. I'd assued the "Remove pre-process syntax" part would generate malformed JS because of things like mismatched braces.

e.g. do we not have some places that do horrible things like:

if (x) { 
#if foo
  xxx
}
#else
  yyy
}
#endif

[kleisauke]

Indeed, I noticed that too. I think the easiest way to fix this is to do it via:

@@ -3,6 +3,9 @@ if (x) {
   xxx
 }
 #else
+#if 0 // STRIP_PREPROCESS
+else {
+#endif
   yyy
 }
 #endif

i.e. just add another always-false preprocessor directive, so that stripping it produces valid JS.

I started experimenting with this in this changeset:
main...kleisauke:preprocess-compat

With that changeset and this strip-preprocess.py script:

Details

#!/usr/bin/env python3

from pathlib import Path
import sys

directives = (
    "if", "ifdef", "elif", "else", "endif",
    "include", "warning", "error", "preprocess",
)

def preprocess_expression(content):
    content = content.strip()

    if content.startswith("asyncIf(") and content.endswith(")"):
        return "async "
    if content.startswith("awaitIf(") and content.endswith(")"):
        return "await "
    if content.startswith("defineI64Param('") and content.endswith("')"):
        return content[len("defineI64Param('"):content.index("')")]
    if content.startswith("maybeExportHeap(") and content.endswith(")"):
        return ""

    return content

for filename in sys.argv[1:]:
    path = Path(filename)
    text = path.read_text()

    # Remove macro and placeholder expressions
    for begin, end in (("{{{", "}}}"), ("<<<", ">>>")):
        while begin in text:
            i = text.index(begin)
            j = text.find(end, i + len(begin))

            if j == -1:
                break

            content = text[i + len(begin):j]
            replacement = preprocess_expression(content)

            text = text[:i] + replacement + text[j + len(end):]

    # Comment out preprocess directives
    lines = []
    for line in text.splitlines(keepends=True):
        s = line.lstrip()
        if s.startswith("#") and any(s[1:].startswith(d) for d in directives):
            line = f"//{line}"
        lines.append(line)

    path.write_text("".join(lines))

I can then run:

$ find src -type f -exec python3 strip-preprocess.py {} +
$ npx @biomejs/biome lint --only=lint/style/useTemplate --write --unsafe --json-parse-allow-comments=true ./src
Checked 135 files in 58ms. Fixed 59 files.

[kleisauke]

I forgot to mention that another nice benefit of doing this is that it also catches invalid JS behind preprocessor directives that aren't always enabled. For example:

emscripten/src/lib/libwebgl.js

Line 3227 in 4b49a33

dbg(`Shader source after preprocessing: ${source}`;

emscripten/src/lib/libwebgl.js

Line 3256 in 4b49a33

dbg(`Shader source after removing layout location directives: ${source}`;

emscripten/src/lib/libwebgl.js

Line 3317 in 4b49a33

dbg(`Shader source after removing layout binding directives: ${source}`;