Earlier this year I presented "Miser's CASE". One thing it lacked
was a "flow through" mode such as found in C's switch statement.

This feature (which some consider a "mis-feature") can easily
be emulated with an addition of a function I call CONTINUE.

But first some background...

When a match occurs in a CASE statement, statements within
the corresponding OF..ENDOF clause are executed, after which
program flow passes straight to ENDCASE .

( 2 -- )
CASE
1 OF ." One " ENDOF
2 OF ." Two " ENDOF
3 OF ." Three " ENDOF
4 OF ." Four " ENDOF
....
ENDCASE

In the above example "Two" is printed. Program flow then passes
to ENDCASE.

Let's now insert a CONTINUE :

On Jun 9, 12:47 am, "Ed" invalid.com> wrote:

"mark4" mailcity.com> wrote in message news:3abd53f7-a53c-491a-924a-d79c3613c05d@v1g2000pra.googlegroups.com...

I'm the last person who wants to make forth look like C :)

Where is the complication? When one looks at the compile
and run-time code, it couldn't be simpler.

Admittedly TEST2 was a poor (and inefficient) example.
A C programmer could probably come up with something
more convincing!

CONTINUE isn't meant to turn CASE into a C switch.
It is a feature that's been available to the classical Forth
CASE but apparently so far overlooked.

Consider it an extra tool in the toolbox - one that costs
nothing to implement.

On Jun 9, 12:47 am, "Ed" invalid.com> wrote:

The real test of this construct is whether it can be used to write
Duff's device, an interleaving of C's while and switch constructs to
implement an efficient partially unrolled I/O port bitbang loop on an
arbitrary length of data.

Here is Tom Duff's own K&R C code for reference:

Ian Osgood quirkster.com> wrote:

Yes, but if it can be used to write Duff's device, would that be a
pass of the test, or a failure?

Andrew.

"theme 2" of http://www.forth.org.ru/~mlg/lections/choose.htm

0 value cs-size
:noname [ depth to cs-size ] ahead [ depth cs-size - to cs-size ]
then ; drop

( n "text}" -- )
: times{ [char] } parse rot 0 ?do 2dup 2>r evaluate 2r> loop 2drop ;
immediate

: cs->r [ cs-size ] times{ postpone >r } ; immediate
: cs-r> [ cs-size ] times{ postpone r> } ; immediate

This definition of CONTINUE is not what I would expect from its
description.

"m_l_g3" yahoo.com> wrote in message news:a7244667-8b19-4037-b760-950e235838ef@56g2000hsm.googlegroups.com...

Yes. Perhaps the CS wordset needs expanding.

"Ian Osgood" quirkster.com> wrote in message
news:94e34c91-c1dd-423b-9a00-7c6032196cfb@y22g2000prd.googlegroups.com...
On Jun 9, 12:47 am, "Ed" invalid.com> wrote:

The real test of this construct is whether it can be used to write
Duff's device, an interleaving of C's while and switch constructs to
implement an efficient partially unrolled I/O port bitbang loop on an
arbitrary length of data.

Here is Tom Duff's own K&R C code for reference:

On Jun 11, 9:20 pm, "Ed" invalid.com> wrote:

On 12 Jun, 18:54, Ian Osgood quirkster.com> wrote:

[...]

An interesting challenge, I think This code does what you want
and I think it is in ANS Forth and therefore portable.

(I'm still using Google Groups as I haven't got round
to changing yet so I'm sorry if the "quoted printables"
upsets anyone's news reader)

It must be the most horrible code I have written in Forth but
it should be quite efficient as all the CS-ROLLing takes place
at compile time.

I've used string evaluation to get the send code in-line,
POSTPONE could be used but a lot of them would be needed.