ACQ unified search too limiting

I think one thing that's missing from the above explanation is that the kind of search that has a problem is, "show me lineitems where invoice id is NOT 3."

Here's a patch that works for me, but wants a little more testing so we can be reasonably sure that it doesn't cause any regressions for other queries:

http://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/collab/senator/lp1031535

I have discovered odd behaviour in the invoice search in 2.3.0 that I think may be related to this bug.

In the invoice search interface if you do the search "PO-State is on-order" with the check boxes for "Sort by title" and "Limit to Invoiceable Items" unchecked you get no results. If you check the boxes for either "Sort by title" and "Limit to Invoiceable Items" then you get results.

I see very similar behaviour when I perform the search "SL-Org Unit is YCL-RH". When "Sort by title" and "Limit to Invoiceable Items" are unchecked very different results are returned than if either box is checked.

It makes sense that "Limit to Invoiceable Items" would change the results, but I expect to get the same results (just in a different order) whether "Sort by title" is checked or unchecked.

I can confirm that with Lebbeous' patch a search on "invoice id does not equal" now returns un-invoiced items as well as invoiced items, whereas before it would only return invoiced items.

Jennifer, just to be clear, the search only displays the first 10 results. You have to use the page navigation links along the top to see other pages of results. If you missed the links (which don't stand out), it might appear that you are getting different results when they are really just sorted to a different page. (If you are using the links, please disregard :)

Jennifer, can you confirm if the issue you reported was a result of paging or if the search was in fact returning different results when sorting by title (taking all pages of results into account)?

Setting to incomplete as we await Jennifer's response.

Since we've been waiting over 7 months, I'm going to un-Incomplete this, Triage-ify it, and ask for more eyeballs. Let's clean up ACQ! :)

Worked for me in testing, and the code (while pretty dense overall) eventually made sense to me. Since a LEFT JOIN is effectively the same as an INNER JOIN if we have a non-null WHERE condition on the table being joined anyway, I couldn't think of a case where changing to a LEFT JOIN would actually hurt this interface.

Pushed to master, rel_2_4, and rel_2_3. Thanks, all!