Bug #10388: Operator precedence problem in multiple assignment (massign)

• Status: Rejected
• Assignee: matz (Yukihiro Matsumoto)
• Target version: -
• ruby -v: ruby 2.2.0dev (2014-10-13 trunk 47904) [x86_64-freebsd10]
• Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN
• Tags: unfixable, parser

[knu (Akinori MUSHA)]

I understand it wouldn't be easy to fix this, but since I happened to find it here it goes.

• a, b = c = 1, 2 is currently taken as a, b = (c = 1), 2; I'd expect it to be taken as a, b = (c = 1, 2).

• a, b = *c = 1, 2 is currently taken as a, b = *(c = 1), 2; I'd expect it to be taken as a, b = *(c = 1, 2) or even a, b = (*c = 1, 2).

• a, b = c, d = 1, 2 is currently taken as a, b = (c), (d = 1), 2; I'd expect it to be taken as a, b = (c, d = 1, 2).

Should they be fixed/changed or not, issuing a warning would be greatly helpful.

[jeremyevans0 (Jeremy Evans)]

I tried working on this a couple weeks ago and I don't believe the current LALR(1) parser can support it. Attempting to modify the parser to support the behavior you desire leads to many shift/reduce or reduce/reduce conflicts. It is possible that switching from the default LALR(1) parser to a GLR parser (which bison also supports) may allow for the behavior your desire, but I'm not sure what the ramifications of that are. It's also possible there is a way to introduce this behavior with the existing LALR(1) parser, and I am just not aware of it, as I do not have much experience in this area.

[hsbt (Hiroshi SHIBATA)]

• Tags set to unfixable, parser

[matz (Yukihiro Matsumoto)]

We are not going to change the behavior. We may warn (with -W) if we see the simple assignments on the right-hand side of multiple assignments.

Matz.

[matz (Yukihiro Matsumoto)]

• Status changed from Open to Rejected