Re: How to add pseudo vector types

[Óscar Fuentes]

Clément Pit-Claudel <cpitclaudel@gmail.com> writes:

> On 7/26/21 4:05 PM, Óscar Fuentes wrote:
>> Think of an Org file with some code blocks. It makes no sense to
>> expose the whole Org file to TS and, I guess, it would just
>> complicate things for no benefit. On that scenario, it might make
>> sense to deal with the code blocks as independent entities instead of
>> parts of something else.
>
> Isn't this example actually in favor of *not* narrowing before giving
> the buffer to TS? Consecutive org code blocks often build upon each
> other, so you'd want to give the whole buffer to TS and restrict its
> analysis to just the code blocks (multiple disjoint ranges), a results
> that you couldn't achieve with narrowing.

I don't know from where you got your "often" :-) Of course it is a
possibility, but I've seen plenty of Org files containing loosely
related code snippets (and those who build on each other tend to be
written on dynamic languages, which benefit a lot less from static code
analysis.)

As far as my personal experience goes, I very much prefer that each
block is treated indepently, because my Org files contain code recipes
for specialized tasks, bug reproducers, multiple variations of
experiments, etc. They are not related at all.

So it is desirable to support both modes well.