[RFC] Fix the Tenary Operator -- Please!? Please?

It's not a feature though - it's a bug. It's just one of those bugs that
the fixing of which may break something.

TL;DR -1 for 7, and -1 in general.

It's not really a bug, it works according to the defined behaviour, the
issue is that some (I think most) people disagree with the correctness of
the defined behaviour.

The rules are quite clear that this cannot go into 7 at this point, and no
exceptions can be made to this rule otherwise there may as well not be a
rule at all.

On the more personal level of my own opinion, I don't believe there is
enough benefit in fixing this to do it at all. It will likely break a lot
of codebases and slow down the adoption of whichever version introduces the
"fix", and all it does is give people the ability to do something which
some (and again I think most) people would agree is a horrible thing to do
from a readability and maintainability perspective.

It also suffers from a similar issue to the "API cleanup" idea that come up
again and again - yes, some arguably poor decisions were made a long time
ago which have left us with inconsistencies and that favourite phrase of
proponents of these ideas, "cognitive overhead". Fixing the inconsistencies
only serves to add cognitive overhead. People may not like how it works,
but at least they know how it works - and anyone writing "run anywhere"
code would still do what they do now, which in my experience is generally
write the ternary chain with brackets to disambiguate it, then realise it
looks horrible and change it to an elseif tree.

On Thu, Mar 26, 2015 at 6:31 PM, Rowan Collins rowan.collins@gmail.com

wrote:

On 26 March 2015 20:54:12 GMT, Michael Morris tendoaki@gmail.com
wrote:

Per PHPsadness...

http://phpsadness.com/sad/30

Since 7 is allowed to have BC breaks this would be the time to fix
this.

I'll let someone with more seniority actually write this up - but
please,
please fix this - it's a very long standing annoyance.

I used to think this was a major nuisance, but recently thought about it
and decided I've never once wanted to do this - the result is really ugly
code whatever the associativity.

What I've always been annoyed by is the precedence of the operator -
having to add brackets to mix it with string concatenation, etc - which
it
turns out to is the same in all sorts of languages.

Regards,

Rowan Collins
[IMSoP]

What I've always been annoyed by is the precedence of the operator -
having to add brackets to mix it with string concatenation, etc - which it
turns out to is the same in all sorts of languages.

It is the ''all sorts of languages'' that is key here. The point is that
PHP
associativity for ?: is different from other languages and it is that that
confuses and leads to bugs. What is right/wrong is not as important as all
others doing it the other way.

--
Alain Williams
Linux/GNU Consultant - Mail systems, Web sites, Networking, Programmer, IT
Lecturer.
+44 (0) 787 668 0256 http://www.phcomp.co.uk/
Parliament Hill Computers Ltd. Registration Information:
http://www.phcomp.co.uk/contact.php
#include <std_disclaimer.h>

--

The feature freeze thing kinda rings hollow for me, as well, because we
were debating this fairly recently before the feature freeze and we still
got pushback from people who feared BC breakage. This is a bug (whether
we're calling it one or not, it is a bug) and it needs to be fixed. PHP
does handle it differently, yes, but the different way in which it's
handling it is wrong. At least, that seems to be the prevailing opinion
among devs.

If we don't put this in 7.0 because of the feature freeze, it needs to go
into the next minor version if the RFC passes. We should've pulled the
trigger on this before the feature freeze and we shouldn't have to wait
until 8.0 because of that.

--Kris

What I've always been annoyed by is the precedence of the operator

• having to add brackets to mix it with string concatenation, etc -
  which it turns out to is the same in all sorts of languages.

It is the ''all sorts of languages'' that is key here. The point is
that PHP
associativity for ?: is different from other languages and it is that
that
confuses and leads to bugs. What is right/wrong is not as important as
all
others doing it the other way.

This reply is competely out of context to the message you quote. Yes, I know that the associativity of the operator is wrong. My contention is that in practice, few people ever try to associate multiple instances of the operator, so few bugs result from it not working.

I contrasted this with precedence, particularly the precedence of an expression mixing the ternary operator with string concatenation, which is a much more common thing to do. But here, PHP is doing the "right" thing, if measured by other languages doing the same.

Personally, I'd be in favour of making the operator non-associative, thus breaking obviously if people try it. Changing it now would result in people using it in PHP 7, then running the code under PHP 5 and having obscure bugs to fix.

Regards,

Rowan Collins
[IMSoP]