stream-sqlite
Python function to extract all the rows from a SQLite database file concurrently with iterating over its bytes, without needing random access to the file.
Note that the SQLite file format is not designed to be streamed; the data is arranged in pages of a fixed number of bytes, and the information to identify a page often comes after the page in the stream (sometimes a great deal after). Therefore, pages are buffered in memory until they can be identified.
Installation
pip install stream-sqlite
Usage
from stream_sqlite import stream_sqlite
import httpx
def sqlite_bytes():
with httpx.stream('GET', 'http://www.parlgov.org/static/stable/2020/parlgov-stable.db') as r:
yield from r.iter_bytes(chunk_size=65_536)
for table_name, pragma_table_info, rows in stream_sqlite(sqlite_bytes(), max_buffer_size=1_048_576):
for row in rows:
print(row)
Recommendations
If you have control over the SQLite file, VACUUM;
should be run on it before streaming. In addition to minimising the size of the file, VACUUM;
arranges the pages in a way that often reduces the buffering required when streaming. This is especially true if it was the target of intermingled INSERT
s and/or DELETE
s over multiple tables.
Also, indexes are not used for extracting the rows while streaming. If streaming is the only use case of the SQLite file, and you have control over it, indexes should be removed, and VACUUM;
then run.
Some tests suggest that if the file is written in autovacuum mode, i.e. PRAGMA auto_vacuum = FULL;
, then the pages are arranged in a way that reduces the buffering required when streaming. Your mileage may vary.