[FP] Avoid linting for $$ with &convert

[evantypanski]

Attributes on a field with &convert make $$ mean something different. $$ refers to before the conversion, self.field refers to after.

Obligatory note: There's more nuance, but I almost see it as a spicy bug. If you run the example and print both self.x and $$, you'll see they'll be the same. But if you try swapping them out in a &requires, they'll be different. That seems wrong to me, so I didn't add that logic in. Maybe worth an issue

BTW this is where it was triggering in spicy Redis, where the replacement makes Spicy not compile the code because that's before a type conversion:

https://github.com/evantypanski/spicy-redis/blob/8101fd33f1c3b3e627e6032741c3c42126790670/analyzer/resp.spicy#L108

[bbannier]

Thanks, this makes sense. Regarding your example

   num_elements: RedisBytes &convert=$$.to_int(10) &requires=self.num_elements <= int64(MAX_SIZE);

I am not sure this is exactly a bug. The Spicy docs document this as

Inside field attributes and hooks, $$ refers to the just parsed value, even if it’s not going to be directly stored in the field. The value of $$ is writable and may be modified.

I agree that the interaction with a following &requires is probably surprising, but it seems to work as designed (we disallow multiple &convert calls). I just quickly checked whether this could misbehave with if $$ is written but this also seems to work.

[evantypanski]

I am not sure this is exactly a bug. The Spicy docs document this as

I'm not too sure either for this case, but I was referring to the fact that $$ refers to something different in &requires (before the &convert) versus in the block (after the &convert). My intuition is $$ should always refer to the pre-conversion value, even in the block. Instead, $$ seems to refer to 2 different things for the same field depending on where it's called. That confused me a fair bit, but it behaves consistently from what I can tell.

[bbannier]

I am not sure this is exactly a bug. The Spicy docs document this as

I'm not too sure either for this case, but I was referring to the fact that $$ refers to something different in &requires (before the &convert) versus in the block (after the &convert). My intuition is $$ should always refer to the pre-conversion value, even in the block. Instead, $$ seems to refer to 2 different things for the same field depending on where it's called. That confused me a fair bit, but it behaves consistently from what I can tell.

Ah, that's not good. Would you mind filing a Spicy ticket for that?