sqlite> .schema
CREATE TABLE stories (
	_id         INTEGER PRIMARY KEY AUTOINCREMENT,
	storyId     TEXT UNIQUE,
	title       TEXT,
	link        TEXT UNIQUE,
	published   TEXT,
	updated     TEXT,
	image       TEXT,
	summary     TEXT,
	month       TEXT
);

CREATE TABLE storyTags (
	_id			INTEGER PRIMARY KEY AUTOINCREMENT,
	_storyId	INTEGER,
	_tagId		INTEGER,
	relevance	REAL,
	FOREIGN KEY(_storyId) REFERENCES stories(_id),
	FOREIGN KEY(_tagId) REFERENCES tags(_id)
);

CREATE INDEX stories1 ON stories ( published ASC, _id );

CREATE INDEX storyTags1 ON storyTags ( _tagId ASC, _storyId ASC );

CREATE INDEX storyTags2 ON storyTags ( _storyId ASC, _tagId ASC );



Get all stories (without duplicates) that have one or more of the
following tagIds: 870, 2012, 21395, ordered by publish date,
oldest story first:

SELECT _id, storyId, title, link, published, updated, image, summary, month
FROM stories
WHERE _id IN (SELECT _storyId FROM storyTags WHERE _tagId IN (870, 2012, 21395)
ORDER BY published;

is faster than joining stories and storyTags. EXPLAIN QUERY PLAN
uses indexes for both selects, but in the joined version only
on the storyTags table - it's actually doing a scan on the stories table.