syntax confusion: 1.+2

[StefanKarpinski]

This sort of confusion keeps cropping up: https://groups.google.com/forum/#!topic/julia-users/aycsd2iAktE. Two possibilities seem possible:

• Eliminate 1. as a short syntax for 1.0;
• Allow 1. but make ambiguous cases like 1.+2 a syntax error.

Related: #18714 – in that it also relates to making potentially ambiguous cases errors.

[ararslan]

+1 for eliminating 1.. That change might be mildly annoying for some people, but it's much clearer to write floating point literals as 1.0 anyway.

The second point, allowing 1. but making 1.+2 a syntax error, seems like more trouble than it's worth.

[tknopp]

given that . is extensively used in Julia I would also vote for eliminating 1.
Making it a syntax error adds new special cases to the parser, which increases the complexity of a (potential) Julia grammar.

[JeffBezanson]

Disallowing 1. seems too strict to me. Many languages allow that syntax, and getting rid of it because of a corner case seems like throwing out the baby with the bathwater. The problems with 1.+2 (either as it is now, or if we add a special case error for it) are much rarer and less significant than the use of 1. as a literal.

[carlobaldassi]

I agree with Jeff even though I personally dislike 1., as I'm sure a lot of people would be very annoyed by not being able to write it/having to change their code. Simply disallowing ambiguous cases should be enough I believe.

[StefanKarpinski]

At this point everyone seems to agree that 1.+2 should be a warning, so let's agree to that. For what it's worth, Ruby doesn't seem to allow 1. as a syntax – probably because the language allows calling methods on integers. C, C++, Python and Perl allow 1.. I'm not sure what R does since I don't have it installed.

[JeffBezanson]

R accepts it, as do octave/matlab, bc, and most lisps/schemes. The syntax is nearly universal.

[mauro3]

Whilst we're at it, are there any problems with omitting the leading zero: .1 which should be taken care of too?

[StefanKarpinski]

There seem to be a lot fewer issues with that syntax.

[StefanKarpinski]

For 0.6 this should be made a deprecation warning or an error, but this change is pretty minor and low impact, so I'm leaving this as up for grabs in case someone wants to do it.

[akaysh]

Hi @StefanKarpinski ,
I am a beginner, can I work on this issue?
Please help me how to get started.
Thanks

[StefanKarpinski]

@akaysh: anyone who wants to can work on it (no need to ask permission), but I suspect this is not a particularly easy issue. It requires some deep hacking on Julia's parser, which is written in Scheme. But if you think that sounds doable, you're welcome to give it a shot.

[Sacha0]

are there any problems with omitting the leading zero: .1 should be taken care of too?

There seem to be a lot fewer issues with that syntax.

(Perhaps due to dot-ops presently having the dot on the left rather than right?)