Remove the DO keyword from PARSE

[rebolbot]

Submitted by: Ladislav

The DO keyword is totally unused and even the bug report for it was written by a user who did not try to write a dialect at all and was just testing the PARSE behaviour.

The fact that the feature is totally unused is a very good proof for me that the feature is gratuitous.

I do not want to give any more arguments since no other argument can have the power of this one. I can only add that this is not the first time I am expressing this opinion and I have been proven correct by the weight of the time that passed since I said it for the first time. Just for the reference I restate that I did not plan to use this feature, do not know about anybody who did, and I am not going to change my mind about it.

Also for the record, there are features I would like to see in PARSE, like the CASE feature (#1898 e.g.), but this one does not make it to the list of the useful ones.

^{CC - Data [ Version: r3 master Type: Wish Platform: All Category: Parse Reproduce: Always Fixed-in:none ]}

[rebolbot]

Submitted by: Gregg

Can you clarify in the description that this tickets is to remove DO from PARSE, not REBOL itself, just to avoid confusion? I know it's marked as the 'parse category, but I missed that when I first read the ticket, having not read the previous ticket.

[rebolbot]

Submitted by: Ladislav

Done, but Rebol (I mean the Do dialect) does not have a DO keyword.

[rebolbot]

Submitted by: BrianH

It was a surprise to me that the keyword had been implemented at all yet, so it's not a surprise that it hadn't been used (especially given that its most important feature hasn't been implemented, see #2083 for details). However, there are better arguments against its use than that it hasn't been used.

More importantly, it's a bad fit for its intended use, mostly by being a PARSE keyword in the first place.

• You can't easily change PARSE keywords since PARSE is monolithic.
• If you don't need the feature then you can't get rid of it, for much the same reason.
• It has to integrate with the rest of the behavior of PARSE, without that being critical for its task.
• The code of PARSE is trickier to work on than a standalone solution would be.
• I've heard the argument that the inclusion of this operation makes PARSE slower. This seems plausible.
• It being included in PARSE inhibits the development of better alternatives.
• There are minor security implications for PARSE simply by having this operation. Very minor, but still.

There are also reasons that aren't really related to it's being a PARSE keyword, but are still important:

• It's not recursive: You can't call back to expressions in your dialect from the DO dialect expressions.
• It's not compilable, so your dialect process can't be sped up.
• It's really only useful for dialects that mix Rebol evaluation rules into non-Rebol evaluation rules, without a clear syntax delineation between the two (like the parens in PARSE, ironically enough). This is usually a bad idea to do in a dialect for usability reasons - it's better to either go for full DO evaluation, something that is mostly DO plus some minor extras (like object spec's treatment of setwords or module block keywords), or something completely different and unified.

Added to all this, now that Rebol is open-source, the need to make dialects without native code is less important, since using native code is easier than it used to be. What is more important is having a variety of different approaches to choose from, with different approaches, developed by the community. It's better to go with many optional, add-on (but embeddable) solutions to this problem, rather than something built into PARSE.

[hostilefork]

DO has been removed, but UPARSE supports combinators which do something like it:

https://github.com/metaeducation/ren-c/blob/master/tests/parse/examples/evaluate.parse.test.reb