rjmac

That is an entirely plausible feature. In fact, we do a similar thing already for the tech preview sync jobs, which use the same infrastructure as datasync. It wouldn't be...

The reason was to make control files that contained erroneous fields (particularly ones that were being used by automated jobs) demand a human's attention instead of quietly doing a potentially...

Indeed we do have this! Implemented it on a whim because it was easy and seemed like it might be useful. See [the SoQL spec](https://github.com/socrata-platform/soql-reference/blob/master/doc/soql.tex) for more. I should probably...

@thedodd updated; you're right about being the same feature. I kept my change which made the config values `IpAddr`s rather than `String`s (because it seems right to push that sort...

I've taken a look at [implementing the libpq blocking api in terms of non-blocking calls](https://github.com/rjmac/postgresql-libpq/commit/4667b9efb56bf0bbaafa5733bfc72984ed236102) (without bothering about the Windows issues) to see what would be involved, and I don't...

I used the latest git source for execStart, but it hasn't been modified in a substantial way since 2011 when bidirectional `COPY` was added, sometime between 9.1 alpha 2 and...

From what I can tell by inspection, if `PQsendQuery` returns 0, you need to call `PQflush` and examine _its_ result to find out if it's an actual failure. 0 means...

It does do that, actually, though it's hidden among all the other fully-emulate-blocking-mode noise -- it's in `flushOutput`, which is called right at the top of `execFinish`.

My main worry about doing things this way is less user-friendliness -- libpq-int.h's dependencies are a package install away on most platforms -- than with binary compatibility. The only _really_...

Heh, maybe. GHCJS isn't super important to me anymore, alas, but maybe it'd be useful for someone!