Complete overhaul of DBI interface. DBI functions now apply translation to all incoming SQL, thus making the DatabaseConnector DBI interface an interface to a virtual database platform speaking OHDSISql as defined in
SqlRender. This should aid write once - execute everywhere code development. It also allows using
dropEmulatedTempTables() function can be called on all platforms (it doesn’t do anything on platforms not requiring temp table emulation).
Adding logging at ‘TRACE’ level for performance profiling using the
ParallelLogger package. Use
options(LOG_DATABASECONNECTOR_SQL = TRUE) to enable. The query times can be automatically extracted using the new
inserTable() function now converts logical fields to integer fields (and throws a warning if it does so).
getTableNames() functions return results in lowercase instead of uppercase for consistency with DBI functions.
Throwing informative error if user calls
createConnectionDetails() with credential arguments that cannot be evaluated in a multi-threading setting.
Supporting external tables on RedShift.
Adding support for ODBC connections to Spark.
Querying to Andromeda when using a DBI driver (instead of a JDBC driver) now also uses batching to avoid running out of memory.
insertTable()and all column names require quotes.
Supporting upcoming version of
dbms() function, which is the preferred way to establish the dbms of a connection.
Adding support for Snowflake.
Updating RedShift driver to V18.104.22.168. Fixes error when uploading data.
Fix erroneous dates when inserting dates in a tibble into a SQLite database.
Ensuring errors when inserting data turn into R errors.
Throwing informative error when using
insertTable() on BigQuery without specifying the
connectionStringis empty string (instead of
Fixing ‘DBMS not supported’ error when connecting to Hive.
Fixing error when bulk uploading to Postgress with NULL values.
Fixing warning when automatically converting
numeric in R 4.2.0.
Updating Spark JDBC driver in response to Log4J vulnerability.
Using new Andromeda functions for better abstraction.
More informative error messages when user forgets to provide a connection property (like a password).
oracleTempSchema in DBI functions.
Splitting vignette into two, because many users will only need to now how to connect to their database.
Improved error messaging related to the driver folder.
getTableNames()when the database or schema name contains escaped characters.
32-bit integers are also converted to numeric by default because (a) the output of some SQL functions (e.g. DATEDIFF) is INT on some platforms, BIG_INT on others, and (b) dplyr often throws errors when working with both integer and numeric vectors.
Reusing byte buffer when passing 64-bit integers from Java to R for efficiency.
Adding support for connecting to Spark.
Throw informative error when provided
dbms argument does not match any of the expected values.
getAvailableJavaHeapSpace() function for debugging purposes.
Fixing field type of numeric fields on Oracle when fetching data.
Fixing issues when fetching dates from SQLite (needed casting to numeric in some scenarios before conversion to date in R).
insertTable() when using a
databaseSchema argument on SQLite.
insertTable() for inserting into temp table on RedShift when table name does not start with ‘#’.
tempEmulationSchema argument. Avoids deprecation warning.
Throwing an informative error when the user points the
pathToDriver argument to a file instead of a folder.
JDBC drivers are no longer pre-packaged in the
DatabaseConnectorJars package. Instead, drivers need to be downloaded manually to a user-specified location in the local file system. The
downloadJdbcDrivers() function has been added to make downing drivers easier. The path to the drivers can be set in the ‘DATABASECONNECTOR_JAR_FOLDER’ environmental variable.
Support for 64-bit integers using bit64’s
integer64 type, although this is disabled by default because of the danger of silent errors when using
integer64 in R. Each query function has a
integer64AsNumeric argument, and overall behavior can be set using
options(databaseConnectorInteger64AsNumeric = FALSE).
INT fields are now translated to R integers and back.
ConnectionDetails delays evaluation of sensitive arguments until needed for improved security.
schema argument from
useMppBulkLoad argument of
insertTable() function in favor of
oracleTempSchema argument in various functions in favor of
tempEmulationSchema argument, which can be set via the “sqlRenderTempEmulationSchema” option.
Adding bulk upload mode for PostgreSQL.
Batch mode in
executeSql divides SQL into batches (1000 statements per batch) to avoid running our of Java heap memory.
Removed message that ‘JDBC driver supports batch updates’ when running executeSql in batch mode.
databaseSchema argument to
insertTable for consistency with general OHDSI framework.
Fixed ‘Warning: no non-missing arguments to max; returning -Inf’ when a column only has NA values, or no values at all.
Fixed error when trying to insert data from a tibble.
Error report now includes all SQL in a batch when failing in batch mode.
Dropping support for orphaned ff package.
Adding support for Andromeda package to replace ff.
Error report file name now defaults to errorReportSql.txt to avoid confusion with other error reports.
Enforcing delay if about to exceed query quota on BigQuery.
Deprecating schema argument in connect and createConnectionDetails functions.
Checking table and column names for SQL reserved words when inserting a table.
Switching to Java 1.8 as minimum requirement.
Preparing for DatabaseConnectorJars v1.2.0, which will have JDBC v4.2 drivers (instead of JDBC v4).
Setting append = TRUE when not creating table in insertTable for RSQLite.
On RedShift, converting schema name to lowercase before getting table names to fix empty results.
Fixed insertTable on BigQuery.
Fixing autocommit not supported error on BigQuery.
Preventing scientific notation when bulk uploading to PDW to avoid error.
Fixing null error when calling getSchemaNames for BigQuery.
Not adding ‘#’ prefix when performing insert into RedShift.
Disabling autocommit when sending updates to RedShift to prevent errors with new JDBC driver.
Preventing ‘FeatureNotSupportedError’ from terminating query on platforms that do no support autocommit.
Fixing some errors related to insertTable when determining correct field type for columns.
Adding support for SQLite through RSQLite (mainly for demonstration and testing)
Adding convenience functions renderTranslateExecuteSql, renderTranslateQuerySql, and renderTranslateQuerySql.ffdf
Dropping Starschema BigQuery driver (in favor of newer Simba driver)
Added support for inserting BIGINTs (large integers stored as numeric in R)
Applying CTAS hack to improve insertion performance for RedShift (was already used for PDW)