InlineQuerier: Location of SQLite DB?

FME Configuration: Temporary folder is `c:\temp', set from environment variable `FME_TEMP'

Thanks @david_r, that is indeed what I suspected.  The Journal file (temporary file that stores uncommitted DB transactions) confirms that is the SQLite DB creation method used by InlineQuerier.

Hmmm, well then next is probably for me to submit an idea for InlineQuerier to expose the in-memory option, which is generally programatically simple to do, all you do when calling something like the sqlite3_open()  function is to pass it a file path string argument =  ":memory:" instead of a more conventional folder/file name path like eg. "%FME_TEMP%tempdb.db".

I've tested that the equivalent SQLite FeatureWriter accepts and writes to a file path of ":memory:" so it seems that the underlying technology to do the equivalent in InlineQuerier is already there in FME?  From experience, these RAM instances are lightning fast for query execution and not relying on having a physical SSD or similar to have high performance.

I think that's a great idea, but I agree that it needs to be an option. Unfortunately there are too many servers out there already struggling with limited memory.

Perhaps also post a link to the idea in this thread, so that it's easy to find for those stopping by this question later.

Thanks @david_r, to link this to the Idea: posted here:

https://knowledge.safe.com/idea/104456/inlinequerier-expose-virtual-ram-option.html?

I've also discussed that there is a 3rd option if concerned about RAM availability that SQLite has, that you can also use the same base database creation command to let SQLite take care of creating and managing its own temporary database file. In these instances it uses this file to "page" any in-memory data overflow to disk but still executes the bulk of the SQL against an in-memory representation. This would address the environment you've noted where might want to be more protective of RAM.

https://www.sqlite.org/inmemorydb.html

It kind of what the background to the original question was: SQLite already provides a programmatic way to call temporary databases and you can let it manage the whole temporary instance instead of FME having to manually create and delete its own on-disk file within the code for InlineQuerier. The authors of SQLite already provide this functionality, which in most cases probably would perform better in Transformer speed.