nathan11g

I agree that leaving out the uid is a good workaround, giving you full control over the order things are listed. I'm still suspicious that the developers intend the order listing to depend on uid _and_ query string (or possibly just the existence of query string, I can't tell). It seems like depending only on the uid would be more flexible.

Sorry for the delay, I thought I had email notifications on. Here is an example workflow: http://orbt.io/P1oG.alfredworkflow Train it a while by typing "grigg" and selecting the top item, and then typing "grigg foo" and selecting item 2. For me, it took about 5 of each to exhibit the problem. Once it is trained, then when you type "grigg" you will see the items in one order and as soon as you begin typing an argument, the order will switch.

I am making a workflow that uses a script filter. The keyword is "date", a space is required, and the argument is optional. The script returns xml with 4 <item>s, and the items include uids. The uids correspond to different date formats, and do not depend on the arguments given. I am expecting Alfred to use the uids to put the more recently used date formats at the top. This is mostly true, but with one problem. It appears that Alfred is tracking two sets of usage data, one for the keyword with arguments and another for the keyword without arguments. So if I type "date", it shows at the top the date format that I usually pick when not using arguments, whereas if I type "date [some argument]" it shows me the date format that I pick when using arguments. This causes undesirable shuffling of result rows. If the issue is unclear, let me know and I can add a video or something.