New EEP draft: Pinning operator ^ in patterns

[Raimo Niskanen]

On Fri, Jan 15, 2021 at 11:48:14AM +0100, Eugene wrote:
> On 1/15/21 11:42 AM, Raimo Niskanen wrote:
> > On Thu, Jan 14, 2021 at 06:09:53PM +0100, Eugene wrote:
> >> On 1/14/21 3:58 PM, Raimo Niskanen wrote:
> >>> On Mon, Jan 11, 2021 at 02:50:46PM +0100, Eugene wrote:
> >>>> annotating UNbound variable will take less symbols.
> >>>> far far far less.
> >>>>
> >>>
> >>> Every time you introduce a new variable, it is unbound.
> >>>
> >>> Do you claim that matching an already bound variable is more common?
> >>
> >> not matching,
> >> mentioning.
> >>
> >> i hope you are not trying to introduce another
> >> context-dependant mental mess.
> > 
> > Nope.  I was trying to understand you.
> 
> if you compare occurrences of UNbound vs bound,
> bound is way more often,
> and should be given the shortest notation.
> 
> if you limit your scope to the "match" context
> (whenever a part of an expression is equivalent to
> a left-side of a match)
> you thus create a context-dependency that
> MUST BE KEPT IN MIND OF A READER.

Now I think I understand.

This EEP suggests to add the pinning operator to variables in patterns,
not in expressions.  Therefore it is better to have the operator on unbound
variables that are less common in patterns.

Patterns are, as you say, left of = (the matching operator), in function
clauses, in case clauses, and in list comprehensions left of <-.

The code reader must already today be very aware about what is a pattern
and what is not.  This EEP does not introduce a new requirement on context
awareness for the code reader.


> 
> you win a statistical comparison and can give
> the shorter notation to UNbound mentions of a variable,
> but a reader of your program has to pay some price.

-- 

/ Raimo Niskanen, Erlang/OTP, Ericsson AB