Re: Vim9 import [was Vim9 script feature-complete]

[Bram Moolenaar]

Marvin Renich wrote:

> * Bram Moolenaar <Br...@moolenaar.net> [220103 07:07]:
> >
> > > On 2021-12-30, Bram Moolenaar <Br...@moolenaar.net> wrote:
> > > > If you have remarks on something in Vim9 script that you think should be
> > > > different, speak up now! Soon we'll only make backwards compatible
> > > > changes to avoid breaking existing plugins.
> > >
> > > There was a thread about this on HN a few days ago, and one of the
> > > comments suggested to reverse the import syntax so that the imported
> > > script is named first and the imported entities next. A rationale for
> > > this is that it would help automatic completion.
> >
> > Is there another language that works like this? The current syntax
> > resembles Javascript, which hopefully some users are familiar with.
>
> I have to agree that putting the file being imported first is much
> better.
>
> In Go, by default every imported identifier must be prefixed by the
> package name, much like the current Vim9 "as" clause. Requiring an "as"
> clause (or defaulting to some form of the script name for the value of
> the "as" clause), and requiring that all imported identifiers be
> prefixed by this identifier make the code _much_ clearer and easier to
> follow.
>
> I very much dislike the JavaScript style where importing, by default,
> places the imported names in the current namespace rather than in a
> separate namespace (identified by the "as" clause in Vim9).
>
> Given that even when you are only importing one identifier from a
> script, the whole script must be read and executed, I think the Go way
> is better. Don't bother with the
>
> import MyClass from "myclass.vim"
> import {someValue, MyClass} from "thatscript.vim"
>
> syntax, and only provide
>
> import "myclass.vim"
> import "myclass.vim" as Other
>
> and require use of the namespace prefix:
>
> Other.MyClass
>
> The first case, without the "as" would default to the file name, with
> leading directories and trailing ".vim" removed, and with the first
> letter capitalized. Non-identifier characters and leading digits would
> be removed as well. In the case without "as" above, you would reference
> the imported MyClass identifier as
>
> Myclass.MyClass
>
> With this, every use of an imported identifier is clear and explicit.

I have to agree that the Javascript way is a bit complicated and
verbose. Especially the "* as SomeName" part is not optimal:
import * as funcs from "./subdir/myfunctioncollection.vim"
Now "funcs" is somewhere in the middle of the line.
This form might actually be the most common use.

Not mentioning the items being imported has the disadvantage that it's
harder to find out what is used from the imported script. But that
might not be very important. You can always use "*" on the name to see
what's being used.

File names are often longer, thus the "import as" style can be expected
to be used quite a lot. Example:

import "./subdir/myfunctioncollection.vim" as funcs
funcs.DoSomething()

The main disadvantage is that the prefix must always be used, there is
no way to use an imported item without it. But as you mention, this can
also be seen as an advantage, that it's clear it is not defined in this
script file.

I hope others give there opinion, this is an important choice.

--
$ echo pizza > /dev/oven

/// Bram Moolenaar -- Br...@Moolenaar.net -- http://www.Moolenaar.net \\\
/// \\\
\\\ sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///

[Bram Moolenaar]

> On 2022-01-03, Marvin Renich <mr...@renich.org> wrote:
> > Don't bother with the
>
> > import MyClass from "myclass.vim"
> > import {someValue, MyClass} from "thatscript.vim"
> >
> > syntax, and only provide
> >
> > import "myclass.vim"
> > import "myclass.vim" as Other
> >
> > and require use of the namespace prefix:
> >
> > Other.MyClass
> >
> > The first case, without the "as" would default to the file name, with
> > leading directories and trailing ".vim" removed
>

> I do not think that using a filename as an identifier is a good idea.
> For instance, calling a script 1.vim would automatically make it
> non-importable (without "as").

Since a script needs to use "export" to be able to be imported, having
to use a nice name for the script is clearly needed. The only thing is
that it may be a long name to avoid name collisions, but then the "as
{name}" form can be used to shorten the name.

> I personally find that using an imported name without a prefix (as it is
> currently possible) makes my code terse, and I think that in the limited
> scope a plugin that works well. But I understand that Vim9 scripts might
> have a broader use, such as generic libraries of functions that can be
> used by many scripts. In that context, stricter scoping rules, such as
> in Go, are likely a cleaner approach.
>
> How about always requiring a prefix, but allowing explicit namespace
> pollution? As in
>
> import "myclass.vim" as Other
> use Other # Makes Other.F() available as just F()

Throwing everything from "Other" into the current namespace is going to
cause trouble, because someone may add an item to myclass.vim that
conflicts with what is in your script. Thus extending one script may
break another script, that is bad.

A kind of an alias mechanism would work. Since most of the items are
going to be functions, might as well use a function reference:

final Func = Other.Func

That's an existing mechanism, thus keeps things simple.

It also works for constants and read-only variables:

const MAX = Other.MAX_VALUE

This does not work for variables that can be set, but you should
probably use a setter function for that anyway.

I'm just pointing out what valid pro's and con's are, I'm not suggesting
this means the alternate import syntax is best.

--
hundred-and-one symptoms of being an internet addict:
208. Your goals for the future are obtaining a second Gbit connection
and upgrade your NAS to all SSD

[Bram Moolenaar]

Marvin Renich wrote:

> * Bram Moolenaar <Br...@moolenaar.net> [220103 12:33]:

> > > On 2022-01-03, Marvin Renich <mr...@renich.org> wrote:
> > > > Don't bother with the
> > >
> > > > import MyClass from "myclass.vim"
> > > > import {someValue, MyClass} from "thatscript.vim"
> > > >
> > > > syntax, and only provide
> > > >
> > > > import "myclass.vim"
> > > > import "myclass.vim" as Other
> > > >
> > > > and require use of the namespace prefix:
> > > >
> > > > Other.MyClass
> > > >
> > > > The first case, without the "as" would default to the file name, with
> > > > leading directories and trailing ".vim" removed
> > >
> > > I do not think that using a filename as an identifier is a good idea.
> > > For instance, calling a script 1.vim would automatically make it
> > > non-importable (without "as").
>

> I agree that using the (cleansed) file name is suboptimal, but it was
> the simplest choice. There are a couple other possibilities. One is to
> require the "as" clause.
>
> Another is to do something similar to Go (a language I like, if you
> couldn't tell :-) ). Every Go source file has a package statement,
> «package frob», and when importing, this package identifier, «frob», is
> used as the prefix (if not overridden in the import statement).
>
> Every vim9 script file already has a vim9script statement. You could
> say that in order for the script to be imported the vim9script statement
> must be of the form «vim9script ScriptId» where ScriptID must be a
> capitalized identifier. If you import two different scripts with the
> same script ID, you must use the "as" clause for at least one of them.
>
> Alternatively, you could require either the script ID on the vim9script
> statement or the "as" clause on the import statement.

Adding a script ID adds another mechanism and I don't see enough
advantage in it. It raises questions, such as what happens if two
completely unrelated plugins use the same ID?

Since the import can use a relative file name, a short file name can
work. It's only when using a file name in 'runtimepath' that we can
expect the name to be longer. Thus requiring the use of "as" up front
does not seem necessary.

> > Since a script needs to use "export" to be able to be imported, having
> > to use a nice name for the script is clearly needed. The only thing is
> > that it may be a long name to avoid name collisions, but then the "as
> > {name}" form can be used to shorten the name.
> >
> > > I personally find that using an imported name without a prefix (as it is
> > > currently possible) makes my code terse, and I think that in the limited
> > > scope a plugin that works well.
>

> My opinion is the opposite, here. Even in small, simple scripts, the
> prefix makes the code more readable; there is no question from where the
> identifier came.

Right. Somehow code writers can be very lazy typing things, and then
spend lots of time (possibly much later) figuring out what the code is
doing. Unfortunately I'm not aware of any studies being done on this
(it's more computer art than computer science).

> > > But I understand that Vim9 scripts might
> > > have a broader use, such as generic libraries of functions that can be
> > > used by many scripts. In that context, stricter scoping rules, such as
> > > in Go, are likely a cleaner approach.
> > >
> > > How about always requiring a prefix, but allowing explicit namespace
> > > pollution? As in
> > >
> > > import "myclass.vim" as Other
> > > use Other # Makes Other.F() available as just F()
>

> I like this very much; it works regardless of how the prefix gets
> defined ("as" clause, vim9script statement, or cleansed filename).
>
> I think if I had to pick, I would require the "as" clause. It is
> simple, and doesn't depend on the script author keeping the same script
> ID with newer versions of the script. The author of the script doing
> the importing is required to be in control.

>
> > Throwing everything from "Other" into the current namespace is going to
> > cause trouble, because someone may add an item to myclass.vim that
> > conflicts with what is in your script. Thus extending one script may
> > break another script, that is bad.
>

> This is probably the best reason to not allow blindly importing all
> identifiers from one script into the local namespace of another.

I'm starting to more and more like the idea of the simplistic import. The
main reason is that the Javascript way suggests that it is possible to
import individual items from a script, which in reality the whole script
is read, while some items can't be adressed.

Also since it's possible to do this:

import FuncOne from "thatscript.vim"
import FuncTwo from "thatscript.vim"
import FuncThree from "thatscript.vim"

This doesn't actually load the script three times, but you'll have to do
digging in the help to know that. Thus there is some hidden knowledge.
It can be written as:

import {FuncOne, FuncTwo, FuncThree} from "thatscript.vim"

Which turns it into a one-liner, but adds the need for a syntax that
isn't used anywhere else.

While using:

import "thatscript.vim"

Or:

import "thatscript.vim" as that

Is nice and short, no need for "from". The help for ":import" becomes
much shorter. The implementation will also be much simpler.

The discussion about the need for using the prefix, whether "as" should
be required and other things, seem less important.

--
A scientist is someone who knows exactly how an engine works, but
can't fix his car when it fails to start.
An engineer is someone who knows only some things about an engine, but
can fix his car when it's broken.

[Tony Mechelynck]

I see an important disadvantage: with this system, if in the importing script you don't explicitly rename the namespace, then when reading the source it isn't possible to know that after

        import foobar.vim
       call ScriptID.Function()

the ScriptID is defined by a statement in foobar.vim. If the "implicit namespace" were easily deductible from the script name (as it already is in the autoload mechanism) then you would have, maybe,

        import foobar.vim
        call foobar.Function()

or something similar, which is much more transparent.