Overcoming the DuckDb rapi_startup issue

[fdefalco]

I found an interesting issue when performing DuckDb testing. I had already performed testing using Sqlite in my environment and because of this the call to Eunomia::getDatabaseFile had download the data set and loaded it into a .sqlite file. When you do this you run into the following behavior:

connectionDetails <- DatabaseConnector::createConnectionDetails(
   dbms = "duckdb", 
   server = Eunomia::getDatabaseFile("Synthea27Nj", cdmVersion=cdmVersion),
   pathToDriver = "~/Documents/OHDSI/Drivers"
 )

ParallelLogger::clearLoggers()
Achilles::achilles(
   cdmVersion = cdmVersion,
   connectionDetails = connectionDetails,
   cdmDatabaseSchema = cdmDatabaseSchema,
   resultsDatabaseSchema = cdmDatabaseSchema,
   smallCellCount = 0,
   createTable = TRUE,
   scratchDatabaseSchema = "main",
   sqlOnly = FALSE
)

Connecting using DuckDB driver
An error report has been created at  output/errorReportR.txt
Error: rapi_startup: Failed to open database: IO Error: Extension "/Users/fjd/Library/Application Support/org.R-project.R/R/duckdb/.duckdb/extensions/v0.8.1/osx_arm64/sqlite_scanner.duckdb_extension" not found.
Extension "sqlite" is an existing extension.

Install it first using "INSTALL sqlite".

What appears to be happening is that Eunomia sees the existing sqlite file and attempts to connect to it with the DuckDb driver. However there is an extension that allows DuckDb to use a sqlite file as a source called the sqlite_scanner which is not installed by default however can be installed in a DuckDb session. To resolve this issue the following code can be executed:

library("DBI")
con <- DBI::dbConnect(duckdb::duckdb(), dbdir = ":memory:")
DBI::dbExecute(con,"INSTALL sqlite")

Once installed you can use the DuckDb driver to connect. An alternative is to delete the sqlite file and Eunomia will create it for you from scratch.