Hi enztec,
the use of )COPY in libapl may depend on whether the workspace that
is being copied was )SAVED or )DUMPED. The latter may use the ∇-editor
which functions only if a proper I/O system is available. I can't say whether
that is the case or not in libapl.
Maybe you should )SAVE (and not )DUMP) all workspaces used with )COPY.
In that case the ∇-editor should not be called because functions are created
directly with ⎕FX. The ∇-editor is actually only a front-end that calls ⎕FX
when the function editing is finished with the final ∇.
Best Regards,
Jürgen
On 3/8/23 7:40 PM, enztec@gmx.com wrote:
Hi Jürgen
please go down the email
On Wed, 8 Mar 2023 18:42:03 +0100
Dr. Jürgen Sauermann <mail@xn--jrgen-sauermann-zvb.de> wrote:
Hi enztec,
I don't quite understand what the actual problem is.
I tested your files fns.c and fns.enz.
Everything looks more or less fine, but I noticed e.g.:
if the folllowing line is commented in fns.c then the f1 fns is not created from what the fns.enz )copy puts into some buffer - so the ∇f1 (nabla editor) is clearly being used in the libapl - it is just that the ∇f1 in the fns.enz when )copied in is not recognized like it is in the 'apl fns.enz' and ap[ then )copy fns.enz examples
apl_exec("∇f1")
which produces:
the error is only produced if the fns definition header is uncommented in fns.enz - so 2 fns headers are run with the above apl_exec("∇f1") being the only code that currently actually creates the fns definition header
SYNTAX ERROR+
Tokenizer: No token for Unicode U+2207 (∇)
Input: ∇f1
i think the problem is is that )copy does not properly parse the fns header properly when used in libapl as it does in apl ws
because the fns body and closing ∇ are sitting somewhere waiting for proper fns definition header to be given as it is in the fns.c with a working apl_exec("∇f1") to produce good fns
As a matter of fact, the ∇ which opens the Nable editor is
being detected and processed before the tokenizer is invoked
and therefore the tokenizer will never see it (and complain
the use of the apl_exec("∇f1") in the fns.c clearly disproves this
as above when it does). The ∇-editor is a purely interactive
feature not intended to be used in libapl. I suppose (since I
haven't written libapl) that input chain looks roughly like this:
┌───────┐
│ input │
└───────┘
↓
┌─────┐
┌←yes←←│ ∇ ? │→→no→→┐
↓ └─────┘ ↓
┌────────┐ ┌────────────┐
│ Nabla │ │ apl_exec() │
│ editor │ └────────────┘
└────────┘ ↓
┌───────────┐
│ Tokenizer │
└───────────┘
Instead of messing around with the ∇-editor you should:
1. take your function lines (header and body lines),
2. quote them so that the lines are valid APL literals,
3. concatenate all quoted lines, separated with a blank,
4. prefix the entrie beast with ⎕FX (unquoted).
5. call apl_exec().
See my previous email for an example. I did the concatenation
in steps 2.-4. in APL to simplify the example, but you can easily do
the same in C/C++ before calling apl_exec().
Best Regards,
Jürgen
On 3/7/23 7:01 PM, enztec@gmx.com wrote:
Thanks Jürgen,
I'd like to keep the situation i gave in my post using the ')copy fns.enz' method as i do fns developement first in the apl ws then test it with apl scripting then to the libapl program and using the apl_exec method you gave would not be practicle.
could you give it an analysis as i think this is a bug and fixing it would be a real plus to the apl/scripting/libapl system
thanks
enztec
On Tue, 7 Mar 2023 16:28:26 +0100
Dr. Jürgen Sauermann <mail@xn--jrgen-sauermann-zvb.de> wrote:
Hi enztec,
see below.
On 3/6/23 9:31 PM, enztec@gmx.com wrote:
Hi
it doesn't seem possible to create apl fns with apl_command or apl_exec directly using libapl This premiss seems wrong:
#include <apl/libapl.h>
// compile with: gcc libapl_test.c -L /usr/local/lib/apl -lapl -lstdc++ -o libapl_test
int
main(int argc, char * argv[])
{
init_libapl(argv[0], 0);
apl_exec( "TEXT ← ⊂ 'Z←A SUM B'" );
apl_exec( "TEXT ← TEXT, ⊂ 'Z←A + B'" );
apl_exec( "⎕FX TEXT" );
apl_command( ")FNS" );
apl_exec( "'⎕CR SUM:' (⎕CR 'SUM')" );
apl_exec( "1 SUM 2" );
}
which produces:
eedjsa@server68:~/apl-1.8/src$ ./libapl_test
SUM
⎕CR SUM: Z←A SUM B
Z←A + B
3
Or, even shorter:
apl_exec( "⎕FX 'Z←A SUM B' 'Z←A + B'");
On 3/6/23 9:31 PM, enztec@gmx.com wrote:
Hi
it doesn't seem possible to create apl fns with apl_command or apl_exec directly using libapl
but i can successfully create a llibapl environment with fns and variables with the following setup and workaround
-
2 files fns.enz and fns.c
--
situation 1 : shows good fns created from fns.enz
/usr/local/bin/apl fns.enz
in apl workspace run
f1
f2 not created if it's fns definition header is commented in fns.enz but is created if uncommented
--
situation 2 :
/usr/local/bin/apl
in apl workspace run
)copy fns.enz
f1
f2 not created if it's fns definition header is commented in fns.enz but is created if uncommented
--
situation 3 :
g++ -O2 fns.c -o fns -L /usr/local/lib/apl -lapl
./fns gives my output below and creates fns1.xml fns2.xml and fns3.xml that can hopefully be of use for analysis of what is happening
in fns.c i can use apl_command(")copy fns.enz"); or apl_exec(")copy fns.enz"); with the fns definition header workaround to get working fns f1 and f2
but without the f1 and f2 fns definition headers in fns.c i get no f1 or f2 created
in the fns.enz ∇f1 and ∇f2 fns definition headers don't work as would hope/expect without the f1 and f2 function definition workaround in fns.c
as you can see in the fns.c after the )copy fns.enz if i do a workaround apl_exec("∇f1"); i get good fns f1 and f2
it seems the ∇f1 fns header doesn't work but the bodies of the fns in fns.enz and closing ∇ are in some 'buffer' and get put into the f1 fns when the f1 function header definitions workarounds is done in fns.c
the interesting thing is is that if i comment the f1 fns definition header in the fns.enz the fns are still created by the corresponding workaround line in fns.c but if i leave it uncommented (which is what it would be if it worked) when i run the correspinding fns header workaround in fns.c it gives the syntax error when run but doesn't prevent it from creating the good f1 fns - so it seems the f1 function definition header in fns.enz is doing something but not creating the fns.
i left the fns.enz f1 function header uncommented and the fns.enz f2 function header commented to show the difference
fns f2 is not created in situation 1 or situation 2 if it's fns definition header is commented in fns.enz but commenting does not affect it's creation in situation 3 (libapl) if f2 workaround fns definition header is used in fns.c
the correct order of the functions in fns.enz and the corresponding fns headers workarounds must be maintained to get proper working fns with the correct names
i have been using this workaround successfully but would love to know what is happening and see if there can be a fix
i have added 3 )save commands at 'strategic' points in fns,c to create the fns1.xml fns2.xml and fns3.xml in hope they give some information that can be used to analyze what is happening
thanks,
enztec
---
this is my output from ./fns from libapl situation 3
)wsid
IS CLEAR WS
)copy fns.enz
DUMPED 2023-03-06 12:31:44 (GMT-7)
)wsid
IS CLEAR WS
)wsid fns1
WAS CLEAR WS
)save
2023-03-06 13:11:51 (GMT-7) fns1
∇f1 workaround 1 in fns.c for f1 fns header in fns.enz
SYNTAX ERROR+
Tokenizer: No token for Unicode U+2207 (∇)
Input: ∇f1
)fns
f1
)wsid fns2
WAS fns1
)save
2023-03-06 13:11:51 (GMT-7) fns2
∇f2 workaround 2 in fns,c for f2 fns header in fns.enz
)wsid fns3
WAS fns2
)save
2023-03-06 13:11:51 (GMT-7) fns3
)fns
f1 f2
f1 fns executed
⍴⍕1 2 3 : 5
f2 fns executed
⍴⍎"1 2 3" : 3
---