Hi,

just wanted to ask if you had the chance to consider Steffen's feature request in the mean time.

Thanks you very much
Svend

I am not sure if explicit connection pooling should be used, implicit should be enough.

Replacing hdb-pool with @sap/hana-client should be quite straight-forward otherwise:

1. driver/sap/SapDriver.ts: SapDriver would need to use the methods from @sap/hana-client for connecting/disconnecting, tracing (events), etc. -> check this.driver and this.master.
2. driver/sap/SapDriver.ts: SapDriver.loadDependencies should only load @sap/hana-client.
3. driver/sap/SapConnectionOptions.ts: keep driver and remove hanaClientDriver. pool and poolSize will also need to be removed/changed (pool contains hdb-pool settings) - new options depending on the pooling strategy might need to be added (e.g. { pooling: boolean, maxPoolSize: number, poolingCheck: boolean, connectionLifetime: number, communicationTimeout: number }) - it's true that the users could also pass them to the driver via BaseDataSourceOptions.extra, but that's deprecated.
4. platform/PlatformTools.ts: remove hdb-pool.

Connection pooling is only supported in @sap/hana-client (NAPI), node-hdb (pure JS) does not support it. However, even though the two drivers should in theory be compatible with each other (except for pooling and a few other things), somehow node-hdb did not work for me correctly out-of-the-box (I actually had to patch both typeorm and node-hdb to get it working). So I doubt there is anyone using the alternative client. Worst case, if implicit pooling is used, the driver will ignore the options and a single connection will exist.

Hi,

thank you for your comment. I had a look into it last year but got stuck in the "otherwise" path pretty quickly. That's why I also forgot to update here, sorry.

Currently it doesn't seem like we would be able to do this.

I haven't used pooling directly from hana-client yet, and SAP documentation is as it always is, but looking into the source code of CAP/CDS can be a good inspiration (if you can understand their architecture).

Currently, hbd-pool is still working for me (heavily patched to support multi-tenant logging though), and since this is probably about a week of work, if I ever look into it, it might not be very soon.

Even if I will implement this, I need to test it in a multi-tenant environment for a while to see that things are still fine :)

I was able to use some time to look further into this. Somehow following your instructions (mostly - did not know what to do with the driver object) at least the connection with implicit pooling according to config was working 😄 - I was able to get back some data from a sample HANA I was using.

I put the POC coding at https://github.com/SvendDomdey/typeorm/tree/poc-wo-hdb-pool. But I did only changes at few points and do not really oversee the consistency of the classes/properties here. Also as you also mentioned, thorough testing will be required (there may be many different use cases).