Gotchas while bulk loading CouchDB
So far I did some conversions from legacy SQL databases to CouchDB. I always had a somewhat different approach.
- I used the primary key of the SQL-DB as the Document-Id. This allowed me to import over and over again without fear of duplicate documents.
- I did row-by-row imports instead of a bulk import. It makes debugging easier. I saw between 5-10 inserts per second over an Internet connection. While this is not lightning fast it was fast enough for me. My biggest Database is 600.000 Documents totaling 20GB. row-by-row bloat the database during import so run compaction occasionally. Then again unless your rows are huge 15.000 rows sounds not much.
My importing code usually looks like this:
def main(): options = parse_commandline() server = couchdb.client.Server(options.couch) db = server[options.db] for kdnnr in get_kundennumemrs(): data = vars(get_kunde(kdnnr)) doc = {'name1': data.get('name1', ''), 'strasse': data.get('strasse', ''), 'plz': data.get('plz', ''), 'ort': data.get('ort', ''), 'tel': data.get('tel', ''), 'kundennr': data.get('kundennr', '')} # update existing doc or insert a new one newdoc = db.get(kdnnr, {}) newdoc.update(doc) if newdoc != db.get(kdnnr, {}): db[kdnnr] = newdoc