A constant theme in OHDSI appears to be the desire to support more and more database platforms, as you can see for example in this thread, this thread, and this thread. I just received an e-mail from an enthusiastic contributor interested in adding Snowflake.
The current state is that we have a core of well-supported platforms, including SQL Server, Oracle, PostgreSQL, and PDW. To my knowledge, all our tools run smoothly on these platforms, and for all except PDW we have unit tests with continuous integration in place to ensure (to some extent) that it stays that way.
Then there are other platforms such as Impala and BigQuery that many in the community have contributed to, but which I am not sure are currently truly supported by the entire OHDSI toolstack.
As you can read in the threads I referenced, I’m in favor of limiting the number of platforms we support, instead focusing on delivering maximum reliability on the platforms we do support. Others disagree, and appear to be willing to provide support for other platforms. However, this has the danger of creating a patchwork of support across our tools. For example, I’m unsure CohortMethod runs on BigQuery or Impala, and if it doesn’t, I would not be able to debug it.
If I can’t have my way (supporting only a few platforms well), then at least we should be transparent about the extent to which we support each platform. For this purposed I’ve set up a strawman Wiki page that tries to list what is required to support a database platform, and has names attached to each tool-platform combination. The idea is that that person is responsible for ensuring the specific tool runs on the specific platform.
Let me know where you stand. Is supporting many platforms a bad idea in general? Are you willing to put down your name in one of the empty cells? Could the ATLAS/WebAPI team help clarify what they think is required to support their software?