Can this be \defnadjx{intervening}{declarative regions}{region!declarative}? Similar question for various other uses of \defna below.

Can you use ! in the third argument to \defnadjx (i.e., in the see{} in \indextext)?

I think we'd need some magic to turn !s into , s. I don't know if \see does that for us but it seems relatively straightforward to build.

That said, should we really be indexing "declarative region" under "region" rather than under "declarative region" in the first place?

We provide "a b, see b, a" because it's standard to look up "noun, adjective", so maybe people do look for "region, declarative" even though the generic word "region" is not useful here? (I also haven't checked to see whether the other uses of \defna are "misuses" like this.)

Well, we don't have index entries for "binding, structured" or "operand, unevaluated". I think that's somewhat appropriate: these aren't "adjective noun" so much as an atomic noun phrase.

I think the right thing here is:

• This should be \defnadj{intervening}{declarative region}
• The "region, declarative" index entry should be changed to say "see declarative region" so that it points to this index sub-entry

You mean \defnadjx{intervening}{declarative regions}{declarative region} for this case, right? And are you saying that (\defnadj and) \defnadjx should be changed to always index adjective noun and generate a see for noun, adjective?

Maybe \defnctx; this is really the same thing as \defncontext but for an inline definition rather than a Terms and Definitions definition, so using a similar macro seems reasonable.

Well, the syntax is a bit different, but you're right about the parallel. Do you have a suggestion for \defna?

Well... should \defna{x}{y} be an alias for \defnadjx{x}{}{y}? (I think we probably always want the see entry, so I'm inclined to say yes.) If so, perhaps we could use \defnadjx{x}{}{y} directly rather than introducing a new macro?

\defnadjx{x}{}{y} renders as "x ", which would be problematic at least if punctuation follows. I'm sure we could teach it to not do so; maybe #1 #2\unskip would be enough.

Another idea for \defnctx: use an optional argument: \defn[entity]{attached}. That seems pretty clean. Although now I come to think of it, that might instead be a nice way to get rid of the \...x macros instead.

Maybe:

• \defnx{bodytext}{indextext} -> \defn[indextext]{bodytext}
• \defnc{bodytext}{indexgroup} -> \defnctx[indexgroup]{bodytext}
• \defnadjx{bodyadj}{bodynoun}{indexnoun} -> \defnadj{bodyadj}[indexnoun]{bodynoun}
• \defna{bodyadj}{indexnoun} -> \defnadj{bodyadj}[indexnoun]{}
  ? (Leaving room for \defnadj[indexadj]{bodyadj}{bodynoun} if we ever want that.)

General idea: the text in braces always appears in the document; the text in brackets is only used to form the index entry.

What do you think?

In general, that seems like an improvement: fewer magic names and less in the way of argument-order guesswork. \newcommand supports only leading optional arguments, but we could certainly use a raw \def for your first \defnadj. We still have to deal with emitting a space only if the bodynoun is non-empty. I don't immediately know how to support optional arguments in two different places, but I'm sure something like \expandafter can be used to do it if it ever matters.

There's a standard trick for an optional argument after a mandatory argument: \defnadj would be a two-parameter command with an optional argument that saves its arguments in some scratch space then expands to another two-parameter command with an optional argument.