util: strip whole CSI sequences per ECMA-48

stripVTControlCharacters() recognises a control sequence by a hand
written list of final bytes, inherited from the bundled copy of
ansi-regex. The list covers neither parameter bytes such as `<` and
`:`, used by mouse reports and by sub parameters, nor final bytes such
as `@`, `X`, `d`, `a` and `b`. A sequence that uses one of them
matches only in part, so the rest of it is left in the string rather
than removed.

getStringWidth() strips before measuring and readline places the cursor
from that width, so the leftovers are counted as printable columns.

Add an alternative built from the control sequence structure in
ECMA-48 5.4: any number of parameter bytes (0x30-0x3F), any number of
intermediate bytes (0x20-0x2F), then a single final byte (0x40-0x7E).
Both existing alternatives are kept and the new one is only reached
where they do not match, so a match can be extended but never shortened.

The new alternative carries its own introducer instead of sharing the
existing prefix, because the shared prefix also consumes `;` and the
two competing for the same run made the match quadratic.

Signed-off-by: Luan Taraschi <130802253+luantaraschi@users.noreply.github.com>

Author: luantaraschi

lib/internal/util/inspect.js

@@ -286,12 +286,24 @@ const meta = [
 // License: MIT by Sindre Sorhus <sindresorhus@gmail.com>
 // Matches all ansi escape code sequences in a string
 const ansi = new RegExp(
+  // Sequences that run until a string terminator, such as OSC 8 hyperlinks.
   '[\\u001B\\u009B][[\\]()#;?]*' +
-  '(?:(?:(?:(?:;[-a-zA-Z\\d\\/\\#&.:=?%@~_]+)*' +
+  '(?:(?:(?:;[-a-zA-Z\\d\\/\\#&.:=?%@~_]+)*' +
   '|[a-zA-Z\\d]+(?:;[-a-zA-Z\\d\\/\\#&.:=?%@~_]*)*)?' +
   '(?:\\u0007|\\u001B\\u005C|\\u009C))' +
-  '|(?:(?:\\d{1,4}(?:;\\d{0,4})*)?' +
-  '[\\dA-PR-TZcf-nq-uy=><~]))', 'g',
+  // Control sequences, using the structure ECMA-48 5.4 defines: any number
+  // of parameter bytes (0x30-0x3F), any number of intermediate bytes
+  // (0x20-0x2F), then a single final byte (0x40-0x7E). The list of final
+  // bytes below covers neither parameter bytes such as `<`, used by mouse
+  // reports, nor final bytes such as `@`, so it stops in the middle of the
+  // sequence and leaves the rest in the output.
+  '|(?:\\u001B\\[|\\u009B)[[\\]()#]*' +
+  '[\\u0030-\\u003F]*[\\u0020-\\u002F]*[\\u0040-\\u007E]' +
+  // Everything else the reference implementation matched, including
+  // sequences whose last byte is a digit, such as `ESC 7`.
+  '|[\\u001B\\u009B][[\\]()#;?]*' +
+  '(?:(?:\\d{1,4}(?:;\\d{0,4})*)?' +
+  '[\\dA-PR-TZcf-nq-uy=><~])', 'g',
 );
 
 let getStringWidth;

test/parallel/test-util-stripvtcontrolcharacters.js

@@ -21,6 +21,30 @@ for (const ST of ['\u0007', '\u001B\u005C', '\u009C']) {
   );
 }
 
+// Ref: ECMA-48 5.4. A control sequence is CSI, then any number of parameter
+// bytes (0x30-0x3F), then any number of intermediate bytes (0x20-0x2F), then a
+// single final byte (0x40-0x7E). Every sequence below is well formed and in use
+// by terminals today, and every one of them used to have its tail left behind.
+tests.push(
+  // Parameter bytes other than digits and `;`.
+  ['a\u001B[<35;10;20Mb', 'ab'], // SGR mouse report
+  ['a\u001B[>1;2cb', 'ab'], // Secondary device attributes
+  ['a\u001B[=5hb', 'ab'],
+  ['a\u009B<35;10;20Mb', 'ab'], // Same, with the 8 bit CSI introducer
+  // Final bytes that the reference implementation does not list.
+  ['a\u001B[3@b', 'ab'], // ICH, insert character
+  ['a\u001B[3Xb', 'ab'], // ECH, erase character
+  ['a\u001B[5db', 'ab'], // VPA, line position absolute
+  ['a\u001B[3bb', 'ab'], // REP, repeat preceding character
+  ['a\u001B[2ab', 'ab'], // HPR, character position forward
+  // Sub parameters are separated by `:`, which is a parameter byte as well.
+  ['a\u001B[38:2:255:0:0mb', 'ab'], // Truecolour in its colon form
+  ['a\u001B[4:3mb', 'ab'], // Curly underline
+  // Intermediate bytes.
+  ['a\u001B[2 qb', 'ab'], // DECSCUSR, set cursor style
+  ['a\u001B[!pb', 'ab'], // DECSTR, soft terminal reset
+);
+
 test('util.stripVTControlCharacters', (t) => {
   for (const [before, expected] of tests) {
     t.assert.strictEqual(util.stripVTControlCharacters(before), expected);