The ODB compiler provides a number of mechanisms for transforming automatically-derived SQL names, such as tables, columns, etc., to match a specific naming convention. At the higher level, we can add a prefix to global names (tables and, for some databases, indexes and/or foreign keys) with the --table-prefix option. Similarly, we can specify custom suffixes for automatically-derived index (--index-suffix; default is _i), foreign key (--fkey-suffix; default is _fk), and sequence (--sequence-suffix; default is _seq) names. Finally, we can also convert all the names to upper or lower case with the --sql-name-case option (valid values are upper and lower).

At the lower level we can specify a set of regular expressions to implement arbitrary transformations of the automatically-derived SQL names. If we want a particular regular expression only to apply to a specific name, for example, table or column, then we use one of the --kind-regex options, where kind can be table, column, index, fkey, sequence, or statement. On the other hand, if we want our regular expressions to apply to all SQL names, then we use the --sql-name-regex option.

The interaction between the higher and lower level transformations is as follows. Prefixes and suffixes are added first. Then the regular expression transformations are applied. Finally, if requested, the name is converted to upper or lower case. Note also that all of these transformations except for --table-prefix only apply to automatically-derived names. In other words, if a table, column, etc., name was explicitly specified with a pragma, then it is used as is, without applying any (except for the table prefix) transformations.

The value for the --*-regex options is a Perl-like regular expression in the form /pattern/replacement/. Any character can be used as a delimiter instead of / and the delimiter can be escaped inside pattern and replacement with a backslash (\). You can also specify multiple regular expressions by repeating these options.

All the regular expressions are tried in the order specified with the name-specific expressions (for example, --table-regex) tried first followed by the generic expressions (--sql-name-regex). The first expression that matches is used.

As an example, consider a regular expression that transforms a class name in the form CFoo to a table name in the form FOO:

--table-regex '/C(.+)/\U$1/'

As a more interesting example, consider the transformation of class names that follow the upper camel case convention (for example, FooBar) to table names that follow the underscore-separated, all upper case convention (for example, FOO_BAR). For this case we have to use separate expressions to handle one-word, two-word, etc., names:

--table-regex '/([A-z][a-z]+)/\U$1/'

--table-regex '/([A-z][a-z]+)([A-z][a-z]+)/\U$1_$2/'

See also the REGEX AND SHELL QUOTING section below.


When entering a regular expression argument in the shell command line it is often necessary to use quoting (enclosing the argument in " " or ' ') in order to prevent the shell from interpreting certain characters, for example, spaces as argument separators and $ as variable expansions.

Unfortunately it is hard to achieve this in a manner that is portable across POSIX shells, such as those found on GNU/Linux and UNIX, and Windows shell. For example, if you use " " for quoting you will get a wrong result with POSIX shells if your expression contains $. The standard way of dealing with this on POSIX systems is to use ' ' instead. Unfortunately, Windows shell does not remove ' ' from arguments when they are passed to applications. As a result you may have to use ' ' for POSIX and " " for Windows ($ is not treated as a special character on Windows).

Alternatively, you can save regular expression options into a file, one option per line, and use this file with the --options-file option. With this approach you don't need to worry about shell quoting.


If the input file is not valid C++, odb will issue diagnostic messages to STDERR and exit with non-zero exit code.


Send bug reports to the odb-users@codesynthesis.com mailing list.