In article <nnd$68843db9$0f3297b6@f356aed0e515fe8b>,
Hans Bezemer <the.beez.speaks@gmail.com> wrote:
<SNIP>

Note: this isn't Python or PHP like "easy to use" - where the magician
hides the rabbit behind his back. The effect is perfectly predictable -
and quite intuitive (if I may say so myself).

Note in C, printf() isn't capable of redirection (unless from OS level). >fprintf() is - but how many people use it in combination with stdout
(present company excepted)?

I sold my soul to the c people, and embraced single bytes characters
and redirection.
: EMIT
DSP@ 1 TYPE DROP
;

: TYPE
1 WRITE-FILE THROW
;

(Windows: replace 1 with STDOUT)

What I like is that I can test all errors/exceptions in my Forth
bye 2>&1 .

The integration with the OS-es makes for useful applications. ~/PROJECT/ciforths/ciforth/TOOLS/ALLOCATE: cat printhex.frt
: doit 1 ARG[] EVALUATE HEX . ;
~/PROJECT/ciforths/ciforth/TOOLS/ALLOCATE: printhex 1024;echo
400

[lina -c printhex.frt]

Hans Bezemer

--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Sun, 18 May 2025 18:55:27 +0000, albert@spenarnc.xs4all.nl wrote:

In article <nnd$68843db9$0f3297b6@f356aed0e515fe8b>,

The integration with the OS-es makes for useful applications. ~/PROJECT/ciforths/ciforth/TOOLS/ALLOCATE: cat printhex.frt
: doit 1 ARG[] EVALUATE HEX . ;
~/PROJECT/ciforths/ciforth/TOOLS/ALLOCATE: printhex 1024;echo
400

[lina -c printhex.frt]

In iForth we can redirect to a string:

: ngspice-POSTLUDE ( -- c-addr u )
DRY-RUN ( using TASK-DATA setup by SP>NGSPICE )
|head .paramdata LOCAL mem
<$$ ." * Generated by iSPICE on " .TIME$ ." -- do not edit"
CR
CR ." .control"
CR ." unset ltspice"
CR ." unset dotstep"
CR ." set numthreads=1
CR ." set ispice = " #|params 0<> 1 AND .
CR ." "
." let spiceidx = $SPICEIDX ; "
." let ntasks=" #|tasks ."#"
." let ncpus=" #|cpus ."#"
." let nparams=" #|params ."#"
." let nresults=" #|results ."#"
#|params
IF CR ." compose p values "
#|tasks 0 ?DO CR ." + "
#|params 0 ?DO mem DF@+ TO mem
F.N2+_
LOOP
LOOP
CR #|params 1 > IF ." reshape p[" #|tasks
0DEC.R ." ][" #|params
0DEC.R ." ]"
ENDIF
ENDIF
CR ." let ignore = 0"
CR ." let buf = 0"
CR ." let done = 0"
CR ." let ix = spiceidx"
CR ." compose arg values " #|tasks #|results * .
#|results 0DEC.R ." ix 0 0.0"
CR ." compose status values ncpus spiceidx"

CR ." dowhile ix < ntasks"
#|params 0 ?DO CR #|params 1 > IF ." let buf = p[ix][" I
0DEC.R ." ]"
ELSE ." let buf = p[ix]"
ENDIF
CR ." alterparam " |head .pnames I CSTR[] .$
." = $&buf"
LOOP
#|params IF CR ." mc_source" ENDIF
CR ." " $SIM @+ 1 /STRING TYPE
#|results 0 ?DO CR ." let arg[3] = " I ."#"
CR ." " I (meashead) TYPE
." arg[4] " I (measbody) TYPE
CR ." let ignore=setresult(arg) "
LOOP
CR ." let ix = ix + ncpus"
CR ." let arg[2] = ix"
CR ." write " RAWDATA-NAME
NO-EXTENSION
S" _{$spiceidx}{.raw}"
$+ TYPE
#|tasks IF CR ." set appendwrite" ENDIF
CR ." end"
CR ." let done = setfinish(status)"
CR ." quit"
CR ." .endc"

SCRIPTFILE-NAME DUMP-TO-FILE
CR ." Please manually add the following lines to `" cktfile .$ ." .sp`"
CR
CR ." .LIBRARY basic.lib basicstuff"
CR ." .OPTION savecurrents"
CR ." .INC " cktfile .$ ." .iscript"
CR ; PRIVATE

( This mixes Forth with the SPICE shell script language )

Result:

* Generated by iSPICE on 23:19:11, December 9, 2024 -- do not edit

control
unset ltspice
unset dotstep
set numthreads=1 $ make sure spinit has this too; we can't overrule spinit!
set ispice = 0
let spiceidx = $SPICEIDX ; let ntasks=1; let ncpus=8;
let nparams=0; let nresults=2;
let ignore = 0
let buf = 0
let done = 0
let ix = spiceidx
compose arg values 2 2 ix 0 0.0
compose status values ncpus spiceidx
dowhile ix < ntasks
TRAN 30u 30ms 0 30u
let arg[3] = 0;
meas tran arg[4] AVG @i1[current]
let ignore=setresult(arg)
let arg[3] = 1;
meas tran arg[4] AVG @i2[current]
let ignore=setresult(arg)
let ix = ix + ncpus
let arg[2] = ix
write f:/2jsource_test_{$spiceidx}{.raw}
set appendwrite
end
let done = setfinish(status)
quit
endc

-marcel

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

In article <fce9942b8d6e52702b330d5edcde5b4f@www.novabbs.com>,
mhx <mhx@iae.nl> wrote:

Mime-Version: 1.0
User-Agent: Rocksolid Light
Path: not-for-mail
From: mhx@iae.nl (mhx)
Newsgroups: comp.lang.forth
Organization: novaBBS
Injection-Info: i2pn2.org; logging-data="921598"; mail-complaints-to="usenet@i2pn2.org"; posting-account="J+nubxJRM7ncpF4l6KLO+OONWmFAYJHVJegfwQXJ8vc";
Subject: Re: Back & Forth - 4tH I/O and You
References: <nnd$306ecdcd$714b9ebb@8c19de68479dc08d> <b2a4690cc4cdcb9e39310166d37070ab2a16d8b2@i2pn2.org> <nnd$68843db9$0f3297b6@f356aed0e515fe8b> <nnd$0346461b$62171617@be8ddde8960df4cb>
Content-Transfer-Encoding: 8bit
X-Rslight-Site: $2y$10$OOQsNDELF5uNo3A3Uv7gJOwfOwKYrncFKKKX4OxnyyB9YDRoHEGxG >X-Rslight-Posting-User: 4e0dc1fdad1ead10b39e7eb5db19bf73d73e3ab3
Date: Sun, 18 May 2025 19:10:34 +0000
Message-Id: <fce9942b8d6e52702b330d5edcde5b4f@www.novabbs.com>
Content-Type: text/plain; charset=utf-8; format=flowed >X-Spam-Checker-Version: SpamAssassin 4.0.0

On Sun, 18 May 2025 18:55:27 +0000, albert@spenarnc.xs4all.nl wrote:

In article <nnd$68843db9$0f3297b6@f356aed0e515fe8b>,

The integration with the OS-es makes for useful applications.
~/PROJECT/ciforths/ciforth/TOOLS/ALLOCATE: cat printhex.frt
: doit 1 ARG[] EVALUATE HEX . ;
~/PROJECT/ciforths/ciforth/TOOLS/ALLOCATE: printhex 1024;echo
400

[lina -c printhex.frt]

In iForth we can redirect to a string:

: ngspice-POSTLUDE ( -- c-addr u )
DRY-RUN ( using TASK-DATA setup by SP>NGSPICE )
|head .paramdata LOCAL mem
<$$ ." * Generated by iSPICE on " .TIME$ ." -- do not edit"
CR
CR ." .control"
CR ." unset ltspice"
CR ." unset dotstep"
CR ." set numthreads=1
CR ." set ispice = " #|params 0<> 1 AND .
CR ." "
." let spiceidx = $SPICEIDX ; "
." let ntasks=" #|tasks ."#"
." let ncpus=" #|cpus ."#"
." let nparams=" #|params ."#"
." let nresults=" #|results ."#"
#|params
IF CR ." compose p values "
#|tasks 0 ?DO CR ." + "
#|params 0 ?DO mem DF@+ TO mem
F.N2+_
LOOP
LOOP
CR #|params 1 > IF ." reshape p[" #|tasks
0DEC.R ." ][" #|params
0DEC.R ." ]"
ENDIF
ENDIF
CR ." let ignore = 0"
CR ." let buf = 0"
CR ." let done = 0"
CR ." let ix = spiceidx"
CR ." compose arg values " #|tasks #|results * .
#|results 0DEC.R ." ix 0 0.0"
CR ." compose status values ncpus spiceidx"

CR ." dowhile ix < ntasks"
#|params 0 ?DO CR #|params 1 > IF ." let buf = p[ix][" I
0DEC.R ." ]"
ELSE ." let buf = p[ix]"
ENDIF
CR ." alterparam " |head .pnames I CSTR[] .$
." = $&buf"
LOOP
#|params IF CR ." mc_source" ENDIF
CR ." " $SIM @+ 1 /STRING TYPE
#|results 0 ?DO CR ." let arg[3] = " I ."#"
CR ." " I (meashead) TYPE
." arg[4] " I (measbody) TYPE
CR ." let ignore=setresult(arg) "
LOOP
CR ." let ix = ix + ncpus"
CR ." let arg[2] = ix"
CR ." write " RAWDATA-NAME
NO-EXTENSION
S" _{$spiceidx}{.raw}"
$+ TYPE
#|tasks IF CR ." set appendwrite" ENDIF
CR ." end"
CR ." let done = setfinish(status)"
CR ." quit"
CR ." .endc"
$$>
SCRIPTFILE-NAME DUMP-TO-FILE
CR ." Please manually add the following lines to `" cktfile .$ ." .sp`"
CR
CR ." .LIBRARY basic.lib basicstuff"
CR ." .OPTION savecurrents"
CR ." .INC " cktfile .$ ." .iscript"
CR ; PRIVATE

( This mixes Forth with the SPICE shell script language )

Result:

* Generated by iSPICE on 23:19:11, December 9, 2024 -- do not edit

control
unset ltspice
unset dotstep
set numthreads=1 $ make sure spinit has this too; we can't overrule
spinit!
set ispice = 0
let spiceidx = $SPICEIDX ; let ntasks=1; let ncpus=8;
let nparams=0; let nresults=2;
let ignore = 0
let buf = 0
let done = 0
let ix = spiceidx
compose arg values 2 2 ix 0 0.0
compose status values ncpus spiceidx
dowhile ix < ntasks
TRAN 30u 30ms 0 30u
let arg[3] = 0;
meas tran arg[4] AVG @i1[current]
let ignore=setresult(arg)
let arg[3] = 1;
meas tran arg[4] AVG @i2[current]
let ignore=setresult(arg)
let ix = ix + ncpus
let arg[2] = ix
write f:/2jsource_test_{$spiceidx}{.raw}
set appendwrite
end
let done = setfinish(status)
quit
endc

That doesn't convince I could do this myself.

-marcel

--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Sun, 18 May 2025 22:51:36 +0000, albert@spenarnc.xs4all.nl wrote:

In article <fce9942b8d6e52702b330d5edcde5b4f@www.novabbs.com>,
mhx <mhx@iae.nl> wrote:

[..]

That doesn't convince I could do this myself.

In Forth we can do everything ourselves.

ngspice-postlude is a word from the netlist converter that translates
a standard SPICE netlist to the iSPICE dialect. The control constructs
tailor the script so that it coordinates parallel execution on a 44
core HPZ840. Each core runs ngspice inside an iForth instance. After
the simulations finish, each copy runs it own set of measurements
on its own rawfile. The data is written to shared memory so that
2.5-D plots can be generated.

-marcel

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

In article <078ff0e4ae28526fe1c04fef6474326237ad928d@i2pn2.org>,
dxf <dxforth@gmail.com> wrote:

On 19/05/2025 2:49 am, Hans Bezemer wrote:

...
Second - I've consulted my script, and nowhere do I use the word "redirection". I *do* state, however: "That means I have to *reuse* as many tokens as possible."

What I mean is: why should I change my source to:

D <# #S ROT SIGN #> fid WRITE-FILE

When I can continue to use a perfectly capable: .

In my case that would be: (.) WRITE

where (.) is a kernel function and WRITE is a library function with >appropriate error handling:

123 (.) write invalid handle error ... aborting

Using that library I could define DISK as mentioned previously and do:

DISK 123 . CONSOLE

but as any exception msg would go to the file, redirection seems more
trouble than worth.

Huh? Redirection is not what you do in your program. It is outside
in the OS. It requires only avoiding silly design choices in
your Forth such as channeling all output through a one character
at a time function as a revectorable EMIT.
If you try to redirect via revectoring EMIT, you have already lost.

Groetjes Albert
--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

In article <nnd$460c410a$64092378@222354ccf17f511d>,
Hans Bezemer <the.beez.speaks@gmail.com> wrote:
<SNIP>

However, the FILE wordset isn't simple. It's an entirely new API you
have to get familiar with. It might be easy to map when your underlying >implementation relies on C, it might be easy to map if you have your
own, internal API that happens to fit nicely on that idea (or simply be
very versatile) - but I doubt it's trivial to map if you have to build
it up from the ground.

It is old and actually very simple. Instead of having one channel
(terminaL) you have channels given a small number as identification.
I have replaced KEY with (ACCEPT) . Instead of streaming one key
at a time you get a string, and you don't bother where it is stored.
All hosted OS's do a better job of letting you correcting input
instead of laboriously programming it into you Forth.
One channel is special, aptly called "standard input".
So you get
: QUIT 1 STATE ! BEGIN (ACCEPT) SET-SRC INTERPRET AGAIN ;
(ACCEPT) is getting an edited string via channel 0 (standard in).

You can open channels to various files. A disk operating system
with named files is enormous progress over blocks.
OPEN-FILE READ/WRITE CLOSE-FILE is a compelling interface since
half a century or longer, and there is no proposal in sight to
replace it.

An still then - since it has virtually no relationship with console I/O
(like TYPE, ACCEPT, ., .", etc.) you have to learn like a completely >different language to use it. So IMHO it fails on the "simplicity" >qualification.

So, no. A little ingenuity solves that to great benefit.

I don't think it's too hard to map the Forth FILE word set on the C file
API. I may do that later if I feel like it. Now - and here we curve back
to my idea of the Forth philosophy - C is not Forth. An idea derived
from the C philosophy is inherently un-Forth-like. Local variables are
part of the stack frame concept - and hence inherently un-Forth-like.

You are wrong to call FILE's c-like. It is operating system like.
You can program your Unix in Haskell and the system calls are the
same.
Local Forth-variables are c-like, and are junk. Real program
languages have functions, that have local variables and local functions
that can have local variables and local functions etc.
Compare the functions for numerical integrating function for
"numerical recipees in C" against "numerical recipees in Pascal".

The argument you can "bolt on" all kinds of C facilities is not an
argument for bolting them on, it's merely an example of and a tribute to >Forth's extensibility.

Agreed. Now bolt on a driver to access a modern disk drive.

<SNIP>

Hans Bezemer

BTW, I have no idea what an inch is - let alone seven of them.

Last time a heard it was a measure of distance.
It is not related to the distance of the equator to the north pole,
but inspired on the human thump.

Groetjes Albert.

--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

In article <a3c8d55396e1a48344a236bec05207ef64d14a72@i2pn2.org>,
dxf <dxforth@gmail.com> wrote:

On 20/05/2025 2:13 am, Hans Bezemer wrote:

On 19-05-2025 15:22, dxf wrote:

Those are console functions which traditionally went to serial ports and bear little
resemblance to disk file I/O.  For implementer it's not simpler - it's a layer on top
of DOS functions.  Even as an end-user I'm not sure I'd like it.  How does one handle
seeks, read and writing to the same file etc?  The ANS functions were designed to
handle random access.  Console functions were not.

Originally, yeah - that was all. And then we enter the wondrous world of "feature requests". So a few more calls were added, like delete-file, reposition-file and the likes.

If you read the documentation you will see that some I/O devices are out of bounds. So we devised tiny little macros for that:

#define UDEV(a)    if(((a)<=COUT)||((a)>=MAXDEVS))throw(M4BADDEV)

That's how we handle things around here.

I think we both felt some simplification of file-handling at the application >programmer level was desirable. For me it's been quite a while since I've >used e.g. READ-FILE directly. The ubiquitous 'filecopy' application for me >would look like this:

: OPENFILES ( -- )
cr .ifile ." -> " .ofile
r/o openin r/w makeout cr
;

: (RUN) ( -- )
openfiles
begin pad 128 read dup while write repeat 2drop
closefiles
;

Or in presence of GET-FILE (sc --sc)

( sc1 sc2 --)
: copy-file 2>R GET-FILE 2R> PUT-FILE ;

I believe the 4tH manual illustrates something similar.

I don't have to worry about overwrite checks, user interrupt, error messages >and so on as it's handled under the hood. Was it all necessary? Is it >'anti-Forth'? Folks can decide for themselves.

Groetjes Albert
--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

In article <nnd$677b4eee$3815c1d3@f69c6d707a0f7f81>,
Hans Bezemer <the.beez.speaks@gmail.com> wrote:

On 20-05-2025 04:45, dxf wrote:

I think we both felt some simplification of file-handling at the application >> programmer level was desirable. For me it's been quite a while since I've >> used e.g. READ-FILE directly. <SNIP>
I believe the 4tH manual illustrates something similar.

As I stated before, there were several requirements coming together in
this design. First, the desire to minimize the number of tokens required
for an I/O system. Second, to design an API which fitted (my idea of)
Forth better. Third, one that made a transition from the previous 4tH
I/O system as painless as possible.

So - yes, you got that right: as simple and as intuitive possible. I
must compliment you on your concise analysis - this is almost exactly as
I tend to write 4tH: you open your files, put the handles on the stack,
do your I/O and use those same handles to close the files. It's
literally coded like that in the "CONVERT" template (See YT: >https://youtu.be/73fDmCh9Mio).

I don't have to worry about overwrite checks, user interrupt, error messages >> and so on as it's handled under the hood. Was it all necessary? Is it
'anti-Forth'? Folks can decide for themselves.

In a way, yeah. Opening a files returns an error code - but the rest of
the API - especially the output words, don't. The reason for that is
I've rarely seen errors in that regard. If you can open a file
successfully, you normally can write to it - at least, that's my
experience. Also, closing files very rarely results in an error. That's
why these parts of the API don't return error codes. Note - I'm not
saying it can't happen. I'm just saying I've only seen it happen very
rarely. So why bug the programmer with endlessly checking error codes?

This is Pascal style, endlessly passing error codes down the command
chain.
GET-FILE ( sc -- sc) doesn't fail normally. It is sufficient to
put a THROW after each XXX-FILE word. No error codes in sight.

If you make a mistake you get
S[ ] OK "poepie" GET-FILE
? ciforth ERROR # -2 : No such file or directory

If you allocate 128 Gbyte for dynamic strings, you don't run out
of memory, normally.
Only if you forget to initialize the pool:

S[ ] OK 10 ALLOCATE THROW
10 ALLOCATE ? ciforth ERROR # 50 : NOT ENOUGH MEMORY FOR ALLOCATE

Thank god for the THROW instead of DROP .

That doesn't mean no error detection is performed. It is - but it THROWs
an error when things go wrong. If you wanna catch it, fine. If not, fine
too. The system will take care of matters anyway. As you state - it's up
to the user to decide.

Just my idea. But then you better be consistent and allow error
codes everywhere. The burden to remember which words needs a throw
and what not makes no sense.


<SNIP>

Groetjes Albert
--
Temu exploits Christians: (Disclaimer, only 10 apostles)
Last Supper Acrylic Suncatcher - 15Cm Round Stained Glass- Style Wall
Art For Home, Office And Garden Decor - Perfect For Windows, Bars,
And Gifts For Friends Family And Colleagues.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)