Docs: document how to proper make batch inserts/updates

[abenhamdine]

Hello,

Regularly there some questions about how to make batch inserts/updates and how to escape values safely.
Eg see :
#1400
#1387
#957
#880

I don't find any reference to this topic in the new doc site, but a user of this library can legitimately wonder how to handle this :

• with a multistatements query ? (no since not supporting parameters)
• with multiple queries and flow control library (eg : async.js async.parallel or Promise.all) ?
• with a multistatements query formatted with a third-party formatting module ? (eg like https://github.com/datalanche/node-pg-format, see example How to insert multiple rows with a single query? #957 (comment))
• with a not-so-obvious query like How to insert multiple rows with a single query? #957 (comment) ?
• within transaction or not ?
• and probably an additional question : when to use pg-copy ?

[qas]

👍

[ariofrio]

Any guidance on this? I'm thinking of this in comparison with the way JDBC's PostgreSQL driver works:

2. JDBC batch API. That is PreparedStatement#addBatch() enables driver to send multiple "query executions" in a single network roundtrip. Current implementation, however would still split large batches into smaller ones to avoid TCP deadlock.

   The actions would be much better:

   execute query
   ...
   execute query
   execute query
   execute query
   sync <-- wait for the response from the DB
   

It seems that using Promise.all() might be the equivalent in node-pg. However:

1. Does node-pg client state get corrupted if a client.query() call occurs before another client.query() call has resolved, on the same Client?
2. Does the second client.query() wait until the first client.query() has received a response from the server, or only until the first client.query() has finished sending the query to the server?
3. [Especially if we don't wrap time-overlapping client.query() calls in a transaction], does node-pg queue server responses appropriately to resolve the correct Promise with the correct response?
4. Is there any performance advantage to wrapping the calls in a transaction?

I am planning to migrate a PostgreSQL JDBC microservice that uses 1000-batched inserts to achieve desired performance to node-pg, which is why I am trying to figure this out.

[charmander]

@ariofrio Batched queries – pipelining – is broader than batch insert. pg doesn’t have pipelining yet. I think @brianc is working on it.

I am planning to migrate a PostgreSQL JDBC microservice that uses 1000-batched inserts to achieve desired performance to node-pg, which is why I am trying to figure this out.

You should probably try to do it in fewer queries anyway, like #957 (comment).

[sebastianrothbucher]

case it still helps: temp tables can achieve it - see this comment: #956 (comment)