<I'm migrating a tool called SemiBin to nf-test>. ...
# team-maintainers
s
I'm migrating a tool called SemiBin to nf-test. I also bumped the version, from v1.4.0 to v2.2.0. The previous module used the
SemiBin
command, which no longer exists. From the github:
This will install both the
SemiBin2
command as well (for backwards compatibility), the old
SemiBin
command. For new projects, it is recommended that you exclusively use `SemiBin2`: both commands do the same thing, but `SemiBin2` has a slightly nicer interface.
(and now actually
SemiBin
has been changed to
SemiBin1
in the newest version) Two things here. First, I think that the versions string should stay as
SemiBin: \$( SemiBin2 --version )
, as the developers call it
SemiBin
in their repository, they have just changed the name of the executable. Like
python3
I guess. But I could change it. Second thing is do we have guidelines for when a tool gets a new module for a separate version? I think
MACS2
and
MACS3
are the biggest offenders here personally. Personally I think that we should only have additional versions of the tool if: • There is significant changes to the input/output structure • Enough people need to stick with the previous version of the tool (maybe supports a feature/use-case that is now dropped) • The tool is important enough to warrant this. The flip side of this is that the same applies when adding a new tool, we should generally be wary of adding any tool with a number at the end 😉
j
If it's the same tool location and name is the same, I agree keep it in the same module with the same SemiBin name entry regardless of the executable
For the latter it's sort of hard to say. Like if someone has completely rewritten a tool from scratch, then should be new module. If it's common parlance in the community to use the name with the number then I would be a new module (for findability, e.g. kraken2)
(personal opinions though)