Digest #350 2026-11-10
Contributions
testcase - Test that `word` correctly skips the leading delimiters
Ensure that the specified delimiter—and only that delimiter—is skipped:
t{ char " word """"X" count s" X" compare -> 0 }t
t{ char " word X" count s" X" compare -> 0 }t
Replies
requestClarification - Using a separate stack for `>R`, `R>`, `R@` and friends
If your question is if >R R> R@ have to use the same stack as nest-sys (what calls push and what EXIT pops from some stack) or other system-execution types (see , the answer is no; system-execution types may be on the same stack, they may be elsewhere, or partially here or there, or they can even have a size of 0 cells (e.g., when inlining). So in this sense the topic in the title can be answered: "Yes, you can use a separate stack for >R and friends".
The standard calls the stack used by >R etc. the return stack. So in this sense the answer to "Does a system actually need to place these values on the return stack?" is yes, and consequently, the answer to "Could it use a separate, dedicated stack for this?" is no. But system-execution types can use one or more separate, dedicated stacks for this, which has the same effect, as far as programs and system implementations are concerned.
AFAICS a standard program has no way of determining how many cells on the return stack (if any) a system execution type uses. This should answer your last question.
This posting should answer your questions, so I am closing this RfC. If there is anything still unclear, please reopen it.
Yes, as usual, a standard system can provide non-standard words to do things that have not been standardized. E.g., Gforth provides table (since 0.2) and cs-wordlist (since 1.0) for defining case-sensitive wordlists (the difference being that there are no redefinition warnings for definitions in a table). After a non-standard word or non-standard usage in a program, the system is free to do anything (doing something that does not surprise the programmer is highly recommended, however).
Another possible way in connection with this proposal would be that a standard system could have a non-standard word for making all dictionary searches case sensitive. A system that originally was case-sensitive could do this in order to support programs written for that system that have an environmental dependency on case sensitive dictionary searches.
I'm referring to word lists accessible to the standard program. According to this proposal, can get-order on the standard system return a word list in which the search is case-sensitive?
NB: even after loading a system-specific program, the system may remain standard or become non-standard.
See also 4.2.2 Other program documentation "whether a Standard System exists after the program is loaded".