[Debichem-devel] Bug#1082109: Bug#1082109: Bug#1082109: inchi: support for .sdf (v3000 dialect) is not functional

[Andrius Merkys]

Dear Norwid,

On 2024-09-23 12:29, Norwid Behrnd via Debichem-devel wrote:
> In the more general case of a .mol/.sdf file about models with multiple atoms
> and at least one explicit bond, the reference executable of InChI 1.07.1 is
> able to process data in the v3000 dialect.  Their executable published by
> August 21st, 2024 on
> <https://github.com/IUPAC-InChI/InChI/releases>
> for instance is able to assign pyridine (written by OpenBabel) its InChI
> string.  For documentation, I attach the log of a test run in Linux Debian
> 13/trixie with InChI's reference executable 1.07.1.
> 
> ---
> 
> Issue 52 (on InChI's GitHub repository) addresses a particular case -- .mol or
> .sdf with exactly one explicit atom only.  In this instance, some programs
> write a bound block (for instance the sketcher InChI's demo page currently
> uses, and OpenBabel) though there is no _explicit_ second atom to bound to.
> The file then reads
> 
> ```
> M  V30 BEGIN BOND
> M  V30 END BOND
> ```
> 
> The reference executable of InChI 1.07.1 considers this to be an error.  Other
> programs (for instance Marvin) write a bound block only if there are at least
> two atoms with an explicit bound.  The current implementation in OpenBabel
> 3.1.1 (as packaged for Debian by September 4, 2024) can process either one of
> the two forms thanks to the internal design of OpenBabel.

Right, I am able to reproduce the issue by looking at the log you have 
attached. Most likely the issue is that inchi-1 and inchi_main are two 
different executables, both of them included in the INCHI-1-BIN.zip you 
downloaded upstream. I will check how to build and include inchi-1 in 
Debian package.

Best wishes,
Andrius