#1553 — Change “EscapeSequence 0 [lookahead ∉ DecimalDigit]” to match reality

bug_id: 1553
creation_ts: 2013-06-08 12:40:00 -0700
short_desc: Change “EscapeSequence 0 [lookahead ∉ DecimalDigit]” to match reality
delta_ts: 2014-12-18 06:54:05 -0800
product: Draft for 6th Edition
component: normative change from ES5.x
version: Rev 15: May 14, 2013 Draft
rep_platform: All
op_sys: All
bug_status: RESOLVED
resolution: DUPLICATE
dup_id: 3477
priority: Normal
bug_severity: enhancement
dependson: 3124
everconfirmed: true
reporter: Mathias Bynens
assigned_to: Allen Wirfs-Brock
cc: ["andrebargull", "bugs.ecmascript", "claude.pache", "jorendorff", "lukeh", "mathias", "utatane.tea"]

commentid: 4159
comment_count: 0
who: Mathias Bynens
bug_when: 2013-06-08 12:40:35 -0700

As posted by Yusuke Suzuki (https://mail.mozilla.org/pipermail/es-discuss/2013-June/031149.html):

EscapeSequence ::
CharacterEscapeSequence
0 [lookahead ∉ DecimalDigit]
HexEscapeSequence
UnicodeEscapeSequence

Because of "[lookahead ∉ DecimalDigit]", strictly speaking, "\08" and "\09" are ill-formed StringLiterals.
But engines in the wild (at least V8, SpiderMonkey, JSC) accept this form and recognize it as "\x008".

To bridge the gap, I think it would be better to introduce this form to the spec officially.

commentid: 4170
comment_count: 1
who: Mathias Bynens
bug_when: 2013-06-12 03:17:12 -0700

Added some tests here: http://mathias.html5.org/tests/javascript/string/ (see the last two tests)

commentid: 4171
comment_count: 2
who: André Bargull
bug_when: 2013-06-12 06:31:35 -0700

Behaviour for noctal escapes in strict-mode is different across engines, e.g. the proposed tests only work in V8 and Presto, other engines report a strict-mode error (SpiderMonkey/JSC/IE).

EscapeSequence with leading "0":
- sloppy mode
=> accepted in major engines as "\08" == "\0" + "8"

- strict mode
=> accepted in V8/Presto as "\08" == "\0" + "8"
=> SyntaxError in IE/JSC/SpiderMonkey

EscapeSequence without leading "0":
- sloppy mode
=> accepted in major engines as "\8" == "8"

- strict mode
=> accepted in IE/SpiderMonkey/V8/Presto as "\8" == "8"
=> SyntaxError in JSC

OctalInteger
- sloppy mode
=> accepted in major engines as 08 == 8
=> Note: SpiderMonkey reports warning in shell

- strict mode
=> accepted in IE/Presto/JSC/V8 as 08 == 8
=> SyntaxError in SpiderMonkey

commentid: 8938
comment_count: 3
who: Mathias Bynens
bug_when: 2014-06-14 00:13:52 -0700

Quoting https://bugzilla.mozilla.org/show_bug.cgi?id=957513#c25

> While this is a perfectly valid error to throw, it apparently breaks banks. https://bugzilla.mozilla.org/show_bug.cgi?id=1025107 Throwing a warning and automatically falling back to handling the literal as decimal rather than octal would possibly be preferable (at least for the time being).

Sounds like the spec needs to be changed to match reality after all.

commentid: 8945
comment_count: 4
who: Jason Orendorff
bug_when: 2014-06-16 11:35:48 -0700

That bug is about number literals like 08, not string escape sequences.

But yes, alas, I think we will have to back out the patch we landed in <https://bugzilla.mozilla.org/show_bug.cgi?id=957513>. The spec should do the same.

commentid: 9772
comment_count: 5
who: Claude Pache
bug_when: 2014-08-12 13:53:04 -0700

See also Bug #3124, (which relates to legacy octal escape sequences followed by 8 or 9 in sloppy mode, rather than \0 followed by 8 or 9 in all modes).

commentid: 11095
comment_count: 6
who: Claude Pache
bug_when: 2014-12-18 06:54:05 -0800

I've open Bug 3477 for a definite answer to the more general issue of backslash followed by one ore more digits in string and template literals.

*** This bug has been marked as a duplicate of bug 3477 ***

archives

#1553 — Change “EscapeSequence 0 [lookahead ∉ DecimalDigit]” to match reality