[WIP] Restore compatibility with harmony 2.X

[Bisaloo]

This is still WIP but opening here for visibility and so we can easily link to it while we open issues or feature requests in other projects.

The current changes are the results from a collaboration between @Herbermann and myself.

It's a bit fragile because we resorted to copying the harmony class definition. This is currently the only way lemur code can know what the class looks like and where to access/set specific elements, such as Z_corr. However, this means that even a tiny update to the class definition on harmony side will break this code.

Long-term, it would be easier for lemur if harmony:

• either exposed the class definition in inst/include/, so lemur can LinkingTo harmony and delete the src/harmony_defines.h file created in this PR.
• or exported a setZcorr() method, the same way it is exporting the getZcorr() method. We can then remove src/harmony_defines.h and src/harmony_bridge.cpp created in this PR.

[const-ae]

Hey, thanks for the draft. It actually looks better than I feared this would be. Of course it would be good to have something more stable, but this could be a stopgap solution. Do you actually have an overview when the next Bioconductor deadline is coming up so that the package won't be removed?

I have also been wondering if it might be better to just re-implement the maximum discrepancy clustering. Has any of you looked into this already?

[wolfganghuber]

On 16. Jun 2026, at 23:37, Constantin ***@***.***> wrote: const-ae left a comment (const-ae/lemur#30) <#30 (comment)> Hey, thanks for the draft. It actually looks better than I feared this would be. Of course it would be good to have something more stable, but this could be a stopgap solution. Do you actually have an overview when the next Bioconductor deadline is coming up so that the package won't be removed? I have also been wondering if it might be better to just re-implement the maximum discrepancy clustering. Has any of you looked into this already?

Good point. It could be a useful separate package… And should not be too onerous given modern coding tools. I feel ready to support it as a PI till the end of my life cycle, if one of you writes & submits a good implementation. Wolfgang

…

— Reply to this email directly, view it on GitHub <#30?email_source=notifications&email_token=ABHVETRPQWBI5NZ2O37QBXD5AG43FA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZSGM4DKNZTG432M4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#issuecomment-4723857377>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABHVETQIKJ4IA52LSR4RTGD5AG43FAVCNFSNUABFKJSXA33TNF2G64TZHM3DAOJSG43TQNBTHNEXG43VMU5TINRXGUYDCMRRGI42C5QC>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ABHVETSAJUYPTYIPSYT3EZD5AG43FA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZSGM4DKNZTG432M4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSVGM33PORSXEX3JN5ZQ> and Android <https://github.com/notifications/mobile/android/ABHVETWSTN3TY5MN4CBCLTT5AG43FA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTINZSGM4DKNZTG432M4TFMFZW63VKON2WE43DOJUWEZLEUVSXMZLOOSXGM33PORSXEX3BNZSHE33JMQ>. Download it today! You are receiving this because you are subscribed to this thread.

[Herbermann]

On 16. Jun 2026, at 23:37, Constantin @.***> wrote: const-ae left a comment (const-ae/lemur#30) <#30 (comment)> Hey, thanks for the draft. It actually looks better than I feared this would be. Of course it would be good to have something more stable, but this could be a stopgap solution. Do you actually have an overview when the next Bioconductor deadline is coming up so that the package won't be removed? I have also been wondering if it might be better to just re-implement the maximum discrepancy clustering. Has any of you looked into this already?
Good point. It could be a useful separate package… And should not be too onerous given modern coding tools. I feel ready to support it as a PI till the end of my life cycle, if one of you writes & submits a good implementation. Wolfgang
…

I played around a little but have not followed up much more at the time, because I feared it will take a bit too long. The algorithm itself is well documented in their paper, including some implementation tricks they used. Creating a package that simply does the diversity clustering step shouldn't be too big in scope. I'll look into it!