Plugin not working due to table schema constraints
|Reported by:||mrctrevisan||Owned by:||saigon|
I downloaded bookmarkplugin-r10380.zip, created the egg file and uploaded it to my Trac 0.12 instance (using postgresql 8.3, python 2.5 on linux debian lenny).
I then run "trac-admin trac-env-path update" as suggested by the plugin's post install message and added to my trac user the bookmark related privileges.
When I tried to add a bookmark, the "*" link was apparently inactive (when pressed it did nothing), so I wasn't able to add a bookmark.
After some investigation (i.e. open the "*" link target into a new page and see Postgresql complaining about the "name" column being not null but a null value is supplied) I finally found out that the table "bookmarks" has a pk made of 3 columns:
trac=# \d bookmarks Table "public.bookmarks" Column | Type | Modifiers ----------+------+----------- resource | text | not null name | text | not null username | text | not null Indexes: "bookmarks_pk" PRIMARY KEY, btree (resource, name, username)
but the plugin's code on insert is as follows:
def set_bookmark(self, req, resource): """Bookmark a resource.""" # resource = self.normalise_resource(resource) if self.get_bookmark(req, resource): return db = self.env.get_db_cnx() cursor = db.cursor() cursor.execute('INSERT INTO bookmarks (resource, username) ' 'VALUES (%s, %s)', (resource, get_reporter_id(req))) db.commit()
If I run the following SQL script, the plugin starts working as expected:
alter TABLE bookmarks drop constraint bookmarks_pk; alter TABLE bookmarks add primary key (resource,username); alter table bookmarks ALTER name DROP not null;
The plugin code doesn't seem to be up to date with the table design, or the table design seems too strongly constrained.