marcioAlmada/yay

About the middle dot

marcioAlmada opened this issue · 24 comments

marcioAlmada commented

The middle dot character · was used to draft the DSL.

It's commonly used to denote token type composition:

macro {
    swap ( T_VARIABLE·A , T_VARIABLE·B )
} >> {
    (list(T_VARIABLE·A, T_VARIABLE·B) = [T_VARIABLE·B, T_VARIABLE·A])
}

It's also being used as syntax sugar for layer matching ···, which is very intuitive because it resembles ellipsis ... without clashing with legit T_ELLIPSIS from variadics:

macro { guard (···condition) {···body} } >> {
    if (! (···condition)) {
        ···body
        throw new \GuardError("Guard error.");
    }
}

guard ($something < 3) {
    throw new \InvalidArgumentException('Something is out of bounds.');
}

But turns out it may not be a good idea as there's people using visible spaces on their text editors:

banoy

Some people also reported It's not very easy to type on many keyboards.

But... what to use instead?

marcioAlmada commented

@haskellcamargo in case you get any ideas ^

haskellcamargo commented

Well, I believe the · fits just perfect, because it:

  • Resembles ellipsis ···
  • Remembers composition ·
  • Has no generic context

And I didn't find any other that would do the work like that. Everybody who I talked with said about §, but it is a weird character for this purpose. Realize also that the editor yet differs the whitespace of · by the color (I use visible whitespaces).

marcioAlmada commented

Cool. I polled other people that also use visible white spaces and no one had complains about it too.

This kind of inter punctuation is not so uncommon in natural language, btw:

I also discovered that the · middle dot is used in Golang's plumbing:

Let's keep it as it is for now :)

marcioAlmada commented

Ty for the input.

huanga commented

How about • (0x2022 or &bull;)?

The only argument I have for • is that on Mac OS X, using default U.S. English input, it can be inputted with ⎇ (alt/option, I always get confused when I see that symbol) + 8, which is pretty trivial. Where as I have a hard time finding the key mapping for the middle dot currently used in spec.

assertchris commented

→ is a favourite of mine...

dzuelke commented

How about • (0x2022 or &bull;)?

That's more visible than the mid-dot too!

hikari-no-yume commented

I like @huanga's suggestion because I'm also using OS X and Alt+8 also works on the British layout.

bwoebi commented

· is alt+shift+h for me, alt+8 yields {.

I have no way to directly input the bullet character from keyboard [swiss/french].

marcioAlmada commented

Reopened per community request.

  • Pick a more ergonomic composition identifier (instead of middle dot ·), before the DSL leaves draft state.
hikari-no-yume commented

Oh, I can actually type · with alt+shift+9 (British Mac layout).

@bwoebi: you can type • with alt+shift+$ on the Swiss French and Swiss German Mac layouts, and with alt+ü on the German Mac layout. I found that out by using OS X's Keyboard Viewer and pressing the modifier keys.

  • 👍1
bwoebi commented

• … indeed! Thanks for the hint :-)

marcioAlmada commented

ok, it looks like @ is surprisingly a good candidate that won't create issues for any keyboard layout and won't clash with common valid syntax, AFAIK:

Instead of:

# composition
T_STRING·name

# layering
···some_layer

# parser comb
·ls(
  ·token(T_STRING),
  ·token·(',')
)·label

We would have:

# composition
T_STRING@name

# layering
@@@some_layer

# parser comb
ls@(
  token@(T_STRING),
  token@(',')
)@label

Also no one would write macros with @ ... unless to elide the "shut up" operator itself 😛

macro { @ } >> { }

Looks uglier, but more inclusive. Thoughts?

marcioAlmada commented

@TazeTSchnitzel @dzuelke the problem with @huanga's suggestion IMMO is that we would still have some demographics that wouldn't be able to type • easily (at least on my Pt-BR keyboard, on linux, I still couldn't find the right keys to tap).

hikari-no-yume commented

Using @ would mean you couldn't use that in your own syntax.

huanga commented

I've been thinking more about the • character vs originally proposed · character.

Both of them offers no clean universal way for input. Plus, each carries its own meaning in specific domains (· is used to denote "dot product" in the Mathematics domain, • is generally used to denote a list item), and so we will also need to offer proper escape characters to ensure they're usable in normal context.

On the other hand, @ is pretty much universally the same (shift + 2), and even on other key mappings, due to the wide adoption of the symbol, it should almost always be printed and visible enough to not have to explain. If escaping is unavoidable anyway, and there's already a standard for escaping (\), then mandating the use of an escape to use the character in regular context would heavily tilt the favor towards something that is easy to type like @.

That said, I don't know how much complexity is added by requiring escape character to be used preceding the token, so that will be a separate research assignment for later ;)

huanga commented

Oops, I just realized why @ might not be a good idea: http://php.net/manual/en/language.operators.errorcontrol.php

I'd have to withdraw my support for @ for now, but the sentiment of finding something that is universally easy to type, and adding escape support in regular usage, still stands.

novasdream commented

maybe ¢ or ¬ or something like ._

afaur commented

I am okay with either · or • personally.

I am not sure what text editor everyone uses but I thought it would be worth mentioning here how to setup a shortcut for text expansion.

Vim

  • Edit ~/.vimrc
  • Add
inoremap <S-Tab> ·
  • Press Shift + Tab and it will insert the · character

Sublime Text

Atom

marcioAlmada commented

Redirecting this discussion to RFC #29

icywolfy commented

Needless to say, anything other than normal A-Z0-9 is not universally convenient. (Even those require headache for anyone using a non-Latin language, where you're always switching keyboard layouts, or using caps-lock as a latin-toggle.

Now, on US-English, Windows -- how does one type the middle dot? Would I just have to copy-paste it everywhere?

chris-kruining commented

@icywolfy you could use the alt code (alt+0183 I believe)

assertchris commented

Should this be closed?

marcioAlmada commented

Yes. Closing.