Re: [Rdkit-discuss] Inchi which flavour??

[Paolo T. <pao...@gm...>]

Hi Mike,

please find here a solution which I have just tested and works well on 
both Unix and Windows.

You need to redirect the C++ stderr stream with ctypes around the call 
whose output you wish to grab.

This can be done defining a context manager that uses ctypes:

import os
import sys
import datetime
import ctypes
import io
import tempfile
from contextlib import contextmanager

# Adapted from
# 
https://eli.thegreenplace.net/2015/redirecting-all-kinds-of-stdout-in-python/
if (sys.platform == "win32"):
     kernel32 = ctypes.WinDLL("kernel32")
     # https://docs.microsoft.com/en-us/windows/console/getstdhandle
     C_STD_ERROR_HANDLE = -12
     c_stderr = kernel32.GetStdHandle(C_STD_ERROR_HANDLE)
     c_flush = kernel32.FlushFileBuffers
else:
     libc = ctypes.CDLL(None)
     c_stderr = ctypes.c_void_p.in_dll(libc, "stderr")
     c_flush = libc.fflush

@contextmanager
def stderr_redirector(stream):
     # The original fd stderr points to.
     original_stderr_fd = sys.stderr.fileno()

     def _redirect_stderr(to_fd):
         """Redirect stderr to the given file descriptor."""
         # Flush the C-level buffer stderr
         c_flush(c_stderr)
         # Flush and close sys.stderr - also closes the file descriptor (fd)
         sys.stderr.close()
         # Make original_stderr_fd point to the same file as to_fd
         os.dup2(to_fd, original_stderr_fd)
         # Create a new sys.stderr that points to the redirected fd
         sys.stderr = io.TextIOWrapper(os.fdopen(original_stderr_fd, 'wb'))

     # Save a copy of the original stderr fd in saved_stderr_fd
     saved_stderr_fd = os.dup(original_stderr_fd)
     try:
         # Create a temporary file and redirect stderr to it
         tfile = tempfile.TemporaryFile(mode='w+b')
         _redirect_stderr(tfile.fileno())
         # Yield to caller, then redirect stderr back to the saved fd
         yield
         _redirect_stderr(saved_stderr_fd)
         # Copy contents of temporary file to the given stream
         tfile.flush()
         tfile.seek(0, io.SEEK_SET)
         stream.write(tfile.read())
     finally:
         tfile.close()
         os.close(saved_stderr_fd)

Then, all you need to grab the RDKit warning printed to stderr is use 
the stderr_redirector() context manager around the relevant call, then 
check the grabbed output for relevant content.

For instance, in your example wrap the Chem.MolToInchi() call as follows:

               f = io.BytesIO()
               with stderr_redirector(f):
                   InChi = Chem.MolToInchi(Chem.MolFromSmiles(y))
               grabbed_stderr = f.getvalue().decode('utf-8')
               if ("WARNING" in grabbed_stderr):
                   print("caught: ", grabbed_stderr)

Cheers,
p.

On 09/10/2019 18:10, Mike Mazanetz wrote:
>
> Hi,
>
> Many thanks this, it is very helpful to see some code.
>
> Yes, as it stands, I am yet to get warnings which are seen in stdout 
> being sent to a file, only errors seem to find their way to my files.
>
> Usually Warnings about stereochemistry don’t get captured.  Anyone see 
> this, I’m guessing it’s the same for failed InChI’s too?
>
> Thanks,
>
> mike
>
> *From:*Scalfani, Vincent <vfs...@ua...>
> *Sent:* 09 October 2019 14:40
> *To:* Maciek Wójcikowski <ma...@wo...>; Greg Landrum 
> <gre...@gm...>
> *Cc:* RDKit Discuss <rdk...@li...>
> *Subject:* Re: [Rdkit-discuss] Inchi which flavour??
>
> Hi Macjek and Mike,
>
> If I understand your question correctly, you can specify InChI option 
> parameters when calculating InChIs. Here is an example:
>
> m = Chem.MolFromSmiles('CCC1=CN=C(NC1=O)NC')
>
> Chem.MolToInchi(m)
>
> 'InChI=1S/C7H11N3O/c1-3-5-4-9-7(8-2)10-6(5)11/h4H,3H2,1-2H3,(H2,8,9,10,11)'
>
> Now, try with one of the non-standard options such as FixedH:
>
> Chem.MolToInchi(m,'/FixedH')
>
> 'InChI=1/C7H11N3O/c1-3-5-4-9-7(8-2)10-6(5)11/h4H,3H2,1-2H3,(H2,8,9,10,11)/f/h8,10H'
>
> To answer the question of what happens when the InChI calculation 
> fails, I get an empty string.
>
> m = 
> Chem.MolFromSmiles('[C@H]1([C@H](C1C2[C@@H]([C@@H]2C(=O)O)C(=O)O)C(=O)O)C(=O)O')
>
> Chem.MolToInchi(m)
>
> '  '
>
> There is also an InChI option that can warn on empty structures, and 
> calculate an empty InChI, which I am assuming is supposed to be 
> ‘InChI=1S//’, however, when trying this option I get the same result 
> as above.
>
> Chem.MolToInchi(m,'/WarnOnEmptyStructure')
>
> '  '
>
> I hope that helps.
>
> Vin
>
> *From:*Maciek Wójcikowski <ma...@wo... 
> <mailto:ma...@wo...>>
> *Sent:* Wednesday, October 9, 2019 3:41 AM
> *To:* Greg Landrum <gre...@gm... 
> <mailto:gre...@gm...>>
> *Cc:* RDKit Discuss <rdk...@li... 
> <mailto:rdk...@li...>>
> *Subject:* Re: [Rdkit-discuss] Inchi which flavour??
>
> Mike,
>
> On top of what Greg said what might be particularly useful is an 
> options parameter where you can pass some non default params to InChI 
> call.
>
> śr., 9 paź 2019, 07:22 użytkownik Greg Landrum <gre...@gm... 
> <mailto:gre...@gm...>> napisał:
>
>     Hi Mike,
>
>     The InChI API itself is not exposed. The contents of the
>     module are in the documentation along with some explanations of
>     how to call it:
>
>     http://rdkit.org/docs/source/rdkit.Chem.rdinchi.html
>
>     If something is missing there, please let us know.
>
>     -greg
>
>     On Tue, Oct 8, 2019 at 5:20 PM <mi...@no...
>     <mailto:mi...@no...>> wrote:
>
>         Dear RdKit users,
>
>         I was reading the inchi module docs and I couldn't find
>         methods to call the InChI API.  Are these exposed in RDKit?
>
>         It says the default is the standard Inchi.  What happens when
>         this conversion fails?
>
>         Thanks,
>
>         Mike
>
>         _______________________________________________
>         Rdkit-discuss mailing list
>         Rdk...@li...
>         <mailto:Rdk...@li...>
>         https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
>
>     _______________________________________________
>     Rdkit-discuss mailing list
>     Rdk...@li...
>     <mailto:Rdk...@li...>
>     https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
>
>
>
> _______________________________________________
> Rdkit-discuss mailing list
> Rdk...@li...
> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss