Preamble: unify local variable definitions and selectors?

[mihnita]

No description provided.

[stasm]

@markusicu wrote in #230 (comment):

Also, it seems like a variable definition can have a side effect of its expression leaking into the selection syntax; don't overload unrelated functionalities.

[mihnita]

Comments migrated from the slides

---

Slides comment, Stanisław Małolepszy (@stasm), 1:42 AM Apr 26

Thanks for commenting on this. It makes me realize that we didn't explain this part of the preamble well during the presentation. All statements in the preamble are selectors, and some of them can optionally be bound to a local variable. In the example here, the preamble has a selector bound to $bar, which isn't used because the first and only variant always matches (it has no keys).

---

Slides comment, Mihai Nita (@mihnita), 6:50 AM Apr 26

I don't think that is the problem.
What I'm saying is that there's is no good way to differentiate between a selector and a local variable declaration until very very late.

---

Slides comment, Mihai Nita (@mihnita), 6:56 AM Apr 26

Wait, you mean this does both?
If yes then

• I would expect to see 2 functions, like :plural (with an offset as an option?) and :number, with precision, rounding, etc
• how do you declare a repeatable alias that is not a selector?

[markusicu]

All statements in the preamble are selectors, and some of them can optionally be bound to a local variable.

Yeah, don't. Either define a variable, or specify a selector. Combining the two is very confusing.

[stasm]

The idea to combine selectors and local variable definitions was an attempt to simplify the syntax and the grammar with the aim of making it a good starting point. The reasoning was that something very simple ("the lower bound of complexity", as I called it before) would lend itself better to extension through discussion.

I recognize that this wasn't a good idea. It diverged from what all 3 proposals had envisioned earlier during the sessions with the CLDR TC, and introduced new problems related to larger localization workflows, like leveraging, as @mihnita explained in a meeting a few weeks ago. To summarize @mihnita's points:

• Combining selectors and local variable definitions essentially means that everything in the preamble is a selector, some of them are bound to local variables, and only a subset of them is used for matching.
• This inflates the total number of selectors, which impacts how many variant keys there need to be for each variant. Some of these keys shouldn't be used for matching, which means the would need to be set to the catch-all symbol (*), or elided (see Variants: can there be fewer keys for a given pattern than there are selectors in the preamble? #254).
• These extra keys would in turn impact how leveraging systems can match variants in translations to variants in the source message. Keys are used as variants' "signatures"; adding more keys changes the signature.
• Furthermore, a catch-all in the source language might actually map to a number of known keys in the target language. For example, a gender-neutral sentence in English can map to 2 variants in French.

So let's keep selectors separate from local variable definitions. I think this issue can be resolved (hence the resolve-candidate label) and the exact syntax decided in:

• Enclose selectors so that it's clear what they are #257
• Preamble: If we start in the "code" mode, do we need to delimit the preamble? #251

[romulocintra]

Consensus : Local variables definitions and selectors should be separated

[stasm]

Actually, this isn't fully completed, as the EBNF on develop still doesn't differentiate between local variable definitions and selectors. But let's use #253 to design and track this change.