[erlang-questions] leading underscores on variables versus _

[Ulf Wiger (TN/EAB)]

Matthias Lang
>   f(_x, _x) -> same;
>   f(_,_) -> different.
> 
> Changing _x to behave as many (most?) initially expect would 
> break backwards compatibility, but only for misleading code 
> such as the above...

A bit more seriously, consider the following code from
erl_parse.erl:

yeccpars2(6, ';', __Ss, __Stack, __T, __Ts, __Tzr) ->
 yeccpars1(__Ts, __Tzr, 282, [6 | __Ss], [__T | __Stack]);


So making the change would require a change in yecc at 
the same time. Code wouldn't break unless recompiled with
new underscore semantics, but a yecc that hadn't been 
modified to match.

Of course it begs the question whether there isn't 
other code as well that uses the same convention.

In jungerl, I found that plain_fsm (by me) and 
eradius (by Martin B.) also reference __X type variables.

It's reasonable to assume that there is more code out 
there that would break.

BR,
Ulf W