Environment variables

DBA_DB

Default URL to use to connect to a database. See Connect URLs for details.

TMPDIR

Directory used for writing temporary files.

DBA_REPINFO

Default repinfo.csv file to use when creating a new database, without explicitly providing one.

By default, /usr/share/wreport/repinfo.csv is used.

WREPORT_TABLES

Directory where the file with B table definitions dballe.txt is searched for, together with all the bufr decoding tables needed by wreport.

By default, /usr/share/wreport is used.

DBA_DB_FORMAT

Override the database format to use when creating new databases.

Possible values:

• V7: current stable format (the default)

DBA_EXPLAIN

If present in the environment, then on SQL databases EXPLAIN is run before each nontrivial query, and its output is printed to standard error.

This is used to debug SQL performance problems and help design better queries.

DBA_PROFILE

If present in the environment, its value points to a directory where DB-All.e will write a file for each database session, with a list of all nontrivial queries that are run and their timing information.

This is used to debug performance problems.

DBA_INSECURE_SQLITE

If present in the environment, speed is preferred over integrity when using SQLite databases.

The actual details of what this means can vary. At the moment, the journal is disabled entirely, and interrupting a running program can result in a corrupted database file.

DBA_FORTRAN_TRACE

If present in the environment, the variable points to a file where DB-All.e will write all Fortran API operations that are performed, in a format that can be easily turned into a C++ unit test case.

This is used when debugging unexpected behaviour, to help reproduce bugs in C++ and contribute to regression testing.

DBALLE_TRACE_FORTRAN is the old name for this variable, still supported for compatibility.

DBA_FORTRAN_TRANSACTION

Wrap in a single database transaction everything that happens on a session between idba_begin() and idba_commit(), unless idba_begin() is called with only "read" access levels.

This should make execution faster at least on PostgreSQL and MySQL, and if idba_commit() is not called, like if the program aborts because of an error, then the partial work is rolled back rather than kept in the database.