Re: segfault using string from c?
ok fixed i had to call NimMain(); this fixed everything
PyMODINIT_FUNC
PyInit_fib(void)
{
NimMain();
return PyModule_Create(_module);
}
Re: segfault using string from c?
thx this got the $fpx working now getting a segfault when i do open(), ill investigate further i guess its something related it works perfectly fine when not calling it from c.
Re: segfault using string from c?
Ok, so the problem is likely that Nim runtime is not initialized. Try setupForeignThreadGc() at the beginning of your proc. TBH, i don't remember exactly how this is done...
Re: segfault using string from c?
yes i tried that but if i do $fpx i get a segfault on the $fpx valgrind segfult line the compiled code is at fp = cstrToNimstr(fpx);
Re: segfault using string from c?
What's your C code then? Are you calling this proc from C program or from Nim program? If C program, Nim's GC might not be initialized.
Re: segfault using string from c?
if it helps this is the valgrind output
==9355== Process terminating with default action of signal 11 (SIGSEGV)
==9355== Access not within mapped region at address 0x0
==9355==at 0x665CA1B: lowGauge_3mwQtFaBTgevFrybZdgUNw
(stdlib_system.c:1380)
==9355==by 0x6664A53:
prepareForInteriorPointerChecking_Zqr3EQAAcehIfp3uSECEuQsystem
(stdlib_system.c:1418)
==9355==by 0x6664A53: collectCTBody_zoTIuavlrCHyw3B8MBaLJA_2
(stdlib_system.c:4046)
==9355==by 0x6664BF7: collectCT_zoTIuavlrCHyw3B8MBaLJA
(stdlib_system.c:4115)
==9355==by 0x6664ED8: rawNewObj_QBPo5VW2O56Eeh8hPbQ7Pg
(stdlib_system.c:4210)
==9355==by 0x666512F: newObjNoInit (stdlib_system.c:4226)
==9355==by 0x66651B3: rawNewStringNoInit (stdlib_system.c:4243)
==9355==by 0x6665251: toNimStr (stdlib_system.c:4707)
==9355==by 0x665A7AB: read_csv (fib.c:335)
==9355==by 0x6658C39: first_func (fib.c:22)
==9355==by 0x4E9B9E: PyCFunction_Call (in /home/x/x/venv/bin/python3.5)
==9355==by 0x524413: PyEval_EvalFrameEx (in
/home/x/x/venv/bin/python3.5)
==9355==by 0x52D2E2: ??? (in /home/x/x/venv/bin/python3.5)
and this is my c code
#include "nimcache/fib.h"
#include
#include
static PyObject* first_func(PyObject* self, PyObject* args)
{
char * value;
if (!PyArg_ParseTuple(args, "s", ))
return NULL;
read_csv("/home/x/Downloads/x.csv");
return Py_BuildValue("f", 1.1);
}
Re: Defining procs with a template
Yup, that works.
For anyone looking to spot the difference, here is original signature:
template defineIndex*(name, source: typedesc, extractIt, cmpAB: untyped)
{.immediate.} = ...
When you just remove the immediate, it breaks. Here is the working signature:
template defineIndex*(name, source, extractIt, cmpAB: untyped) = ...
The difference is that _[all](https://forum.nim-lang.org/postActivity.xml#all)
of the arguments are untyped, versus using typedesc for the first two arguments.
@Araq, if you've got a moment (or anyone else that can offer details): why does
this work?
Re: segfault using string from c?
First way is correct, except conversion from cstring to string should be made
with $.
proc read_csv(fpx: cstring): void {.exportc.} =
let fp = $fpx
var s = newFileStream(fp, fmRead)
if s == nil:
echo "file not found"
return
segfault using string from c?
How can you pass a char * to nim, ive been trying for a while but i keep
getting segfault.
if i use:
proc read_csv(fpx: cstring): void {.exportc.} =
let fp: string = cast[string](fpx)
var s = newFileStream(fp, fmRead)
if s == nil:
echo "file not found"
return
then there's a segfault at newFileStream
if i use
proc read_csv(fp: string): void {.exportc.} =
var s = newFileStream(fp, fmRead)
if s == nil:
echo "file not found"
return
then the first 15 chars get truncated, but i still also get a segfault at
newFileStream
i tried calling the function from c with read_csv("/path/to/csv");
if i run the nim file instead of compiling to c it works.
Also i found out that if i hardcode the value i get a segfault here
STRING_LITERAL(TM_xgSOGG8fvQlwPH9bPyp9bbTg_6, "path_here", 36);
thx
Re: Defining procs with a template
Try this prototype: template defineIndex*(name, source, extractIt, cmpAB: untyped) = ...
Re: Nim Dynamic Libraries: exporting functions that return pointer-based types
I have now written a [blog post about this](http://www.bootstrap.me.uk/bootstrapped-blog/nim-for-the-discerning-rubyist) Feel free to ignore the first, introductory, part
Defining procs with a template
Hey folks!
As the immediate pragma is deprecated, I'm updating my packages to remove
references to it. I've got one situation where I'm curious about the best
alternative. It's [this
template](https://github.com/Nycto/RBTreeNim/blob/a64b6a8ad230a8a8f34e7ce1f6564ebf423a5b17/rbtree.nim#L34-L72).
Basically, I'm using a template to define a type with a few procs. When I
remove {. immediate .}, my [test
fails](https://github.com/Nycto/RBTreeNim/blob/a64b6a8ad230a8a8f34e7ce1f6564ebf423a5b17/test/customize_test.nim).
I know I could achieve this using a macro instead, but since all I'm doing is
some basic substitutions, a template feeks a bit more clear. What's the best
way to achieve this?
Re: Arraymancer - A n-dimensional array / tensor library
The only static parts of the Tensor types are the Backend (Cpu, CUDA, ...) and the internal type (int32, float32, object ...). The network topology will be dynamic and using dynamic graphs more akin to PyTorch/Chainer/DyNet than Theano/Tensorflow/Keras. My next step is to build an autograd so people only need to implement the forward pass, backpropagation will be automatic. For this part I'm waiting for VTable. PS: I think NimData is great too, Pandas seems like a much harder beast!
Re: In-Memory Database
Thank you all ! @Stevo, can you pls provide some examples of how to use it ?
Re: Advance Nimble configs?
For anyone facing a similar problem, I centralized my build configuration here: [https://github.com/Nycto/ExtraNimble](https://github.com/Nycto/ExtraNimble) It does (basically) everything I asked for in this thread above, though I had to do a few hacky things to get it working.
