Implementing NAME>INTERPRET

[SirWumpus]

How is NAME>INTERPRET typically implemented? There does not appear to be standard word to mark a word as compile-only (similar to IMMEDIATE). For primitives / built-ins its easy enough, but how is it done for defined words? Do defined words have to be state smart to throw -14 or other if interpreted?

[ruv]

The word name>interpret is insufficiently specified in Forth-2012. An irony is that this word (along with name>compile and find-name) was supposed to replace find, since find is insufficiently specified, but it got into the same problem.

It is supposed that name>interpret returns the xt (execution token) of a word, the same xt that Tick returns for the word (or that find returns in interpretation state).

• Hence, in the general case, to perform the interpretation semantics for a word, this xt should be executed in interpretation state — albeit the glossary entry does not specify this at the moment (therefor you can get an impression that you can execute this xt in compilation state as well as in interpretation state and get the same result — but this is incorrect, since the returned xt may identify state-dependent execution semantics).

In classic single-xt systems this word corresponds to the word name> ( nt -- xt ) that just fetches the xt from a word header.

In dual-xt systems this word returns the primary xt for an nt, which identifies the execution semantics for the word identified by the nt.

• Some people tend to claim that this xt also identifies the interpretation semantics for the word (and the name reflects this idea), but I believe this is incorrect in the general case.

Do defined words have to be state smart to throw -14 or other if interpreted?

In a standard program, user-defined words should somehow analyze state to throw an exception only in interpretation state. Classic single-xt systems do the same. Usually they have a helper word for that, something like:

: ?comp ( -- ) \ throw code -14 if the system is not in compilation state
  state @ 0= -14 and throw
;

[ruv]

I have prepared a proposal to better specify what the results of name>interpret mean.

NB: the standard defines default interpretation semantics (see 3.4.3.2), so they are undefined by the standard only if this fact is explicitly declared as "Interpretation: Interpretation semantics for this word are undefined".