bbv makes it impossible to use $ as a prefix printing notation without parentheses about bbv HOT 15 CLOSED

from bbv.

Is there actually a good reason why $ should be at the extreme level of 0, as opposed to something like 10 (same as function application)? Since both of these are breaking changes, it may make sense to do them together.

from bbv.

Also, some notations are in word_scope, while others don't have an explicit scope, is there any reason for this other than historical?
Not opening word_scope provides some protection from unwanted notations, but probably not what @JasonGross needs, because even if you don't open word_scope, the notations still define the level and associativity in an irreversible way, so they should really be in a separate module.

from bbv.

@andres-erbsen if you put $ at level 10, you have to write f ($4) ($5) instead of f $4 $5, which I think is undesirable. Level 9 is the highest level where f $4 $5 still works. Do you have a particular other notation in mind which would benefit from changing the level of $?

from bbv.

f $4 $5 sounds like an use case that I don't want to argue against, so feel free to pick any level below 10 or keep it at 0 as you wish. Overall, I think avoiding bringing in notations is a better approach than trying to make them compatible -- after all, Haskell people would probably want to put $ at level 100 or something like that to act as ( ... ) $ ( ... )...

from bbv.

I would like to put it at level 9, x at level 10 for usage in fiat-crypto, so that it displays without the parentheses, which is incompatible both with at level 9 and with at level 10 (and with at level 0). This is why I want it to not be exported by default anywhere.

from bbv.

What exactly do you mean by "displays without parentheses"? I can do the following, is this what you meant?

Definition word(sz: nat): Set. Admitted.
Definition natToWord (sz n : nat) : word sz. Admitted.


Notation "$ n" := (natToWord _ n) (at level 0, n at level 0, format "$ n") .

Definition f{sz}(a b c: word sz): word sz. Admitted.

Definition a: nat := 7.
Definition t: word 8 := f $a $5 $(3 + 4).

Print t. (* prints "f $a $5 $(3 + 4)" *)

Whereas currently, this would print as f $ (a) $ (5) $ (3 + 4). I think it would be nice to have $ without parentheses in Word.v too.

from bbv.

Yes, that is what I meant. But, even with this change, I think that both # and $ should be in their own module and not exported by default.

from bbv.

from bbv.

Yes of course we should put the notations inside a module, I was just wondering whether we can get parenthesis-less $ in bbv too because I think that would be nice.

Moreover, it would also be nice if the lemmas in Word.v could still use $, to avoid big refactoring effort. Do you think it would be ok to have the following code inside Word.v? (As in, is $ isolated enough to not spoil what you or anyone else in the future wants to do?)

Module WordNotations.

Notation "$ n" := (natToWord _ n) (at level 0, n at level 0, format "$ n") .

(* more notations... *)

End WordNotations.

Import WordNotations. (* but not exported *)

(* lemmas happily using $ ... *)

from bbv.

I am fine with this. I believe it is similar to the notations in List.v.

from bbv.

@JasonGross would you mind submitting a PR for this?

from bbv.

Bump. Can we get this done? I don't mind submitting a pull request for this.

from bbv.

Yes, please submit a PR @vmurali!
I just merged #9, so you can have the "edit lock" for some time 😉

from bbv.

Note also that putting # at level 0 breaks the infix # notation from the QArith stdlib module.

from bbv.