Add user interrupt check while waiting for query results to be ready.

[zozlak]

Sometimes user wants to terminate a query before a first row of results is available (e.g. it turned out that the query is too complicated and he wants to rewrite it instead of waiting for its completion). So far user interrupts where not checked until PQgetResult() completion so the only option to achieve that was to terminate whole R session or to terminate the query on the database server.

This pull requests adds user interrupts check between PQsendQueryPrepared() and PQgetResult().

It is done by waiting for data on the database connection socket with a timeout (1 second) using glibc's select() and running checkUserInterrupt() if the timeout is hit.

[krlmlr]

Thanks! The new code contains some libpq function calls I'm not familiar with, could we make this optional for now?

[krlmlr]

In this project, we acknowledge in the NEWS.md file. I'll add an acknowledgment for all contributors to r-dbi repos to the upcoming DBI blog post too.

[zozlak]

Moved to NEWS.md in d6dcba9

[zozlak]

The new code contains some libpq function calls I'm not familiar with, could we make this optional for now?

Of course we can. How would you prefer it to be done?

By the way I added a link to the PG documentation in the code and an example of using PQconsumeInput() can be found at https://www.postgresql.org/docs/current/static/libpq-example.html (they don't use PQisBusy() in the example because they are waiting for notifications and not data but the idea is the same)

[krlmlr]

Thanks. I think an argument to dbConnect() would be a good way.

Can you please also resolve the conflicts? I can take care of the NEWS entry.

[zozlak]

Conflicts removed, adding dbConnect() parameter requires a little more time :)

[zozlak]

I've been quite busy last few days, I'll provide the dbConnect() switch during the weekend.

[zozlak]

Any idea how to name the parameter?

[zozlak]

I added it as checkInterrupts but I'm not attached to the name :)

[krlmlr]

Thanks for your PR. I'll tweak myself.

[krlmlr]

Could you please double-check if the code still works as intended after my tweaks? In master now.

[zozlak]

Seems to work fine :)

Just a very minor remark - as you moved the timeval timeout declaration inside the loop in the src/PqResultImpl.cpp:485 it can be written as

timeval timeout = {1, 0};

instead of

timeval timeout = {0, 0};
timeout.tv_sec = 1;

[krlmlr]

Thanks. The code is slightly more expressive in the current state, I'm keeping it.