SYNOPSIS

-- 1. Inspect what changes your deployment would cause to the functions: -- Wrap the SQL your deployment consists of within dollar-quotes and pass it as the first argument to Deploy. The second argument must be NULL. -- The deployment most typically consists of a single CREATE OR REPLACE FUNCTION statement, replacing the existing source code of an existing function, -- but it could of course include statements creating new functions, dropping functions, changing ownership, etc. -- In this step a rollback will be done before the function returns, so the SQL will have no effect, it will only execute it in order to present you with a diff. SELECT Deploy($DEPLOY$ CREATE OR REPLACE FUNCTION Foo() RETURNS BOOLEAN AS $BODY$ DECLARE BEGIN RETURN TRUE; END; $BODY$ LANGUAGE plpgsql VOLATILE; $DEPLOY$, NULL);

-- Example output from Deploy-function: deploy

+-------------------+ | Removed functions | +-------------------+

+---------------+ | New functions | +---------------+

Schema................+ public Name..................+ foo Argument data types...+ Result data type......+ boolean Language..............+ plpgsql Type..................+ normal Volatility............+ VOLATILE Owner.................+ joel Source code (chars)...+ 33

+-------------------------------+ | Updated or replaced functions | +-------------------------------+

MD5 of changes: df62b14663c69887574cc320a2e20d78 (1 row)

-- 2. If the changes were expected and you feel it is safe to commit for real, copy/paste the MD5 sum and pass it as second argument: SELECT Deploy($DEPLOY$ CREATE OR REPLACE FUNCTION Foo() RETURN BOOLEAN AS $BODY$ DECLARE BEGIN RETURN TRUE; END; $BODY$ LANGUAGE plpgsql VOLATILE; $DEPLOY$, 'df62b14663c69887574cc320a2e20d78');