Support incremental BLOB-read in `JDBC3ResultSet::getBinaryStream`

[vdshb]

It was a news for me, that JDBC3ResultSet::getBinaryStream is actually reads all BLOB in one buffer and returns simple ByteArrayInputStream instead of InputStream with incremental BLOB reader.

sqlite-jdbc/src/main/java/org/sqlite/jdbc3/JDBC3ResultSet.java

Line 190 in 8460f63

public InputStream getBinaryStream(int col) throws SQLException {

I believe many JDBC-users don't aware about it either.

It also turns out, that sqlite3 native API make it nearly impossible for current JDBC-implementation to read BLOB incrementally.

I've created an enhancement-request on SQLite-forum to discuss this issue and ask for additional API:

https://www.sqlite.org/forum/forumpost/323f31cffb

I'm surprised someone hasn't created such discussion a decade ago, but later better than never.

This issue is basically a link to SQLite forum discussion an a reminder to implement incremental BLOB reading if something changes on SQLite side.

[vdshb]

It's also my invitation to add arguments/thoughts on SQLite forum discussion, if you have anything to say.

[gotson]

It also turns out, that sqlite3 native API make it nearly impossible for current JDBC-implementation to read BLOB incrementally.

If i understand correctly, what you would like in the JDBC driver is not possible because of sqlite C api, is that a fair statement ?

[vdshb]

@gotson Correct. That's how I see current situation (you might have better expertise, though). Hopefully there will be improvement at some point in the future, and after that improvement of JDBC3ResultSet::getBinaryStream will become an easy change (InputStream implementation with incremental C-calls inside).

[gotson]

In that case i will close this, given we cannot do anything about it until such a function exist in the C library fo Sqlite.

[vdshb]

@gotson I, personally, believe it supposed to be open but with [blocked] badge on it. It's definetly not closed until sqlite team desides to accept/refuse improving their API in this direction.

[sjlombardo]

@vdshb I think you did a great job of outlining a case on the email thread, but it is extremely unlikely that the SQLite team will explicitly accept or refuse the API improvement. If they do implement it, the code will just show up on trunk someday. If that happens then a new issue or pull request would be in order, based on the specific implementation at the time. If they don't, the thread will just go stale. It doesn't make sense to leave this issue open indefinitely when there is no possible way for it to be resolved in isolation.