all synced up

diff --git a/VERSIONING.SKETCH b/VERSIONING.SKETCH
new file mode 100644 (file)
index 0000000..03e6ea1
--- /dev/null
+++ b/VERSIONING.SKETCH
@@ -0,0 +1,30 @@
+Schema versioning/deployment ideas from Jess (with input from theorbtwo and mst):
+1) Add a method to storage to:
+ - take args of DB type, version, and optional file/pathname
+ - create an SQL file, via SQLT, for the current schema
+ - passing prev. version + version will create an sqlt-diff'ed upgrade file, such as
+  - $preversion->$currentversion-$dbtype.sql, which contains ALTER foo statements.
+2) Make deploy/deploy_statements able to to load from the appropriate file, for the current DB, or on the fly? - Compare against current schema version..
+3) Add an on_connect_cb (callback) thingy to storage.
+4) create a component to deploy version/updates:
+ - it hooks itself into on_connect_cb ?
+ - when run it:
+   - Attempts or prompts a backup of the database. (commands for these per-rdbms can be stored in storage::dbi::<dbtype> ?)
+   - Checks the version of the current schema being used
+   - Compares it to some schema table containing the installed version
+   - If none such exists, we can attempt to sqlt-diff the DB structure with the schema
+   - If version does exist, we use an array of user-defined upgrade paths,
+    eg: version = '3x.'; schema = '1.x', upgrade paths = ('1.x->2.x', '2.x->3.x')
+   - Find the appropriate upgrade-path file, parse into two chunks:
+    a) the commands which do not contain "DROP"
+    b) the ones that do
+   - Calls user callbacks for "pre-upgrade"
+   - Runs the first set of commands on the DB
+   - Calls user callbacks for "post-alter"
+   - Runs drop commands
+   - Calls user callbacks for "post-drop"
+ - The user will need to define (or ignore) the following callbacks:
+  - "pre-upgrade", any code to be run before the upgrade, called with schema object, version-from, version-to, db-type .. bear in mind that here any new fields in the schema will not work, but can be used via scalarrefs.
+  - "post-alter", this is the main callback, at this stage, all old and new fields will be available, to allow data migration.
+  - "post-drop", this is the clean-up stage, now only new fields are available.
+