Macroify: str

[lilactown]

str should emit "" + ... code.

I think we still need the core str function when the common (apply str ,,,) is used

[corasaurus-hex]

So there's another option here: string interpolation using template literals.

Note that there's a mild difference between the two syntaxes. Addition would coerce the expression to a primitive, which calls valueOf() in priority; on the other hand, template literal would coerce the expression to a string, which calls toString() in priority. If the expression has a @@toPrimitive method, string concatenation calls it with "default" as hint, while template literals use "string". This is important for objects that have different string and primitive representations — such as Temporal, whose valueOf() method throws.

It seems like we'd want string interpolation since it more closely maps to Clojure's str?

[corasaurus-hex]

I wrote some code to see the differences between the two options: https://gist.github.com/corasaurus-hex/2c03d8abc1c8827b44b59fbdc31a3622

I can't see any difference between them in practice except the ones noted in the paragraph I quoted.

So do we want the behavior of string concatenation or string interpolation? I'd vote for interpolation, both for forwards compatibility and it being the least surprising behavior.

[borkdude]

@corasaurus-hex String interpolation is a different issue: this issue is just about generating more efficient code for (str foo bar) => cherry/str(foo,bar) vs foo + bar.
I think string interpolation should also be a built-in feature that compiles directly to JS string interpolation. We can support this using a reader template #i "foo ${name}". I'll make a different issue for that.

[corasaurus-hex]

The generated code for (str foo bar) can either be "" + foo + bar or it can be `${foo}${bar}` and I had intended to examine which was the better choice.

[borkdude]

I don’t understand why we should emit template code if the user didn’t use a templates string. My expectation would be that a JS runtime will have to do less if you keep it like the user wrote it.

On Wed, 17 Aug 2022 at 13:02, Cora Sutton ***@***.***> wrote: The generated code for (str foo bar) can either be "" + foo + bar or it can be `${foo}${bar}` — Reply to this email directly, view it on GitHub <#10 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACFSBS2ZPP4B67GR5VETRDVZTBD3ANCNFSM56PDHJBQ> . You are receiving this because you commented.Message ID: ***@***.***>

-- https://www.michielborkent.nl https://www.eetvoorjeleven.nu

[corasaurus-hex]

The main reasons, for me:

• template interpolation calls .toString on the elements vs + which calls .valueOf on all the elements. I think that more matches with the spirit of str.
• following from that, calling .toString on the elements makes it forwards compatible with Temporal.
• cljs calls .toString on all the elements, too.

String interpolation via templates is just a means to an end for me, though, what I really care about is using .toString instead of .valueOf and just doing "" + ... will call .valueOf.

[borkdude]

Right. How CLJS compiles (str "foo" "bar") is:

["foo","bar"].join('');

Does that match the .toString behavior? If so, then we could emit that too.

[borkdude]

But then again, we could just handle that in the str function instead instead of inlining this, probably won't matter much.

[lilactown]

I don't have a strong preference. My hunch is that my usage of str would go down in application code if there was a reader conditional for doing string interpolation as in #32.

So instead of doing something like

(str "Hello, " name "!")

I would do

#i "Hello, ${name}!"

The only time I really need str then is for the (apply str ,,,) case I noted in the OP, which can't be expanded to infix anyway. So I think based on that we don't need to macroify it if we go with #32

[borkdude]

Agreed

[corasaurus-hex]

Yes, Array.prototype.join calls .toString and not valueOf