Clarify that .match selectors are not declarations #750
Conversation
eemeli
added
fast-track
Co-authored-by: Addison Phillips <[email protected]>
| > [!NOTE] | ||
| > A _selector_ is not a _declaration_. | ||
| > Even when the same _function_ can be used for both formatting and selection | ||
| > of a given _operand_ | ||
| > the _annotation_ that appears in a _selector_ has no effect on subsequent | ||
| > _selectors_ nor on the formatting used in _placeholders_. |
There was a problem hiding this comment.
| > [!NOTE] | |
| > A _selector_ is not a _declaration_. | |
| > Even when the same _function_ can be used for both formatting and selection | |
| > of a given _operand_ | |
| > the _annotation_ that appears in a _selector_ has no effect on subsequent | |
| > _selectors_ nor on the formatting used in _placeholders_. | |
| > [!NOTE] | |
| > _Expressions_, including _expressions_ that appear as _selectors_, | |
| > have no side effects and are not _declarations_. | |
| > That is, they do not introduce or override _variable_ names. | |
| > This means that the _annotation_ of a _variable_ in a _selector_ | |
| > does not change what is bound to the _variable_, | |
| > even if the same _function_ can be used for both formatting and selection | |
| > of a given _operand_. | |
| > The meaning of a _variable_ reference that occurs within a _variant_ | |
| > depends only on the explicit or implicit _declaration_ of that _variable_. |
Just a suggestion. I think this text is difficult to understand outside the context of the discussions we've been having, and without an example. The same might be true of my suggested text. I guess if it's confusing, people will ask about it during the tech preview feedback period.
There was a problem hiding this comment.
The right hand size of a .local is an expression, so your suggestion might be confusing from that respect?
There was a problem hiding this comment.
Doesn't seem like it to me. An expression is not a declaration; it can appear in a declaration, but not every term that an expression can appear in is a declaration. ("Term" = "subtree of the data model".)
Alternately, the text could begin with something like "Only a .local or an .input declaration can change the meaning of a variable." Though this isn't quite true, as free variables introduce what we call "implicit declarations" someplace. A .match can't contain declarations (not even implicit declarations, since the only place where an "implicit declaration" makes sense is in a context where further names can be introduced) so it's clear from that that nothing within a .match changes the meaning of any variable name.
Every phrasing I can think of is made more complicated by implicit declarations, though.
There was a problem hiding this comment.
I find this suggested text harder to parse than the current succinct text.
We already have the whole surrounding formatting.md which effectively says that same thing that the note does; this is intended to clarify the situation wrt. selectors specifically, and going into details and explanations here doesn't really help with that.
There was a problem hiding this comment.
@catamorphism Would you object to my merging the text as is? We're definitely coming back to the whole "resolved value" space later 😄
There was a problem hiding this comment.
@aphillips I don't object!
There was a problem hiding this comment.
Thanks. Merging now... the end of the 45 train is coming 🚄
Closes #748
Adds a clarifying note about
.matchnot acting as a declaration and updates misleading examples.