CDB file size is 3-5 times original XML!

[duckAsteroid]

The main culprit is that the full XML path keys are massive and repeat a lot!

• could ZIP results?
• could keep a table of path to long IDs and then use the long as key (would mean two phase lookup)

[duckAsteroid]

ZIP works nicely (because the keys repeat) and gets the size back down close to the original XML.

However, the ZIP would need unpacking for read access (since we are using a random access file).

[duckAsteroid]

Storing the path of the keys separately with an id for each would save space...

As an example we could store /world[0]/continent[0]/country[0]/city[0]=1234; then 1234@id'=1234', the downside is that parent/child traversals would require an additional lookup for the key ID (e.g. what is the ID of /world[0]/continent[0]/country[0]/city[0]/name[0]?)