Re: [Gretl-devel] grabbing Alfred

Am 19.08.2013 21:14, schrieb Allin Cottrell: ... Status update: For now it still seems necessary to pre-process the input. I did that with Python in a 'foreign' block within the gretl script, and after some trial-and-error passing around the correctly formatted path strings it works for me as a temporary solution. I will send the code when I have wrapped the stuff in functions as far as possible. Here's how I transformed the Alfred input file (of course I may have overlooked some nicer solutions): As you know, the publication date columns (realtime_start_date and realtime_end_date) come in daily ISO format %Y-%m-%d. But I need to do an earlier/later filtering operation on them. So I stripped the hyphens to get pure numbers. (BTW: I tried to do the same thing with the --tkey column "observation_date" and specify the format there with '--time="%Y%m%d"' [and also "%Y %m %d"], but that produced an error. Expected or a bug?) I also changed the missing value code in the realtime_end_date from "." to the number 99999999 for the same reason, to make numerical comparison possible. In order to avoid this preprocessing, it would be great if it were possible to use a column with daily ISO dates in a filtering comparison directly, as in: '--filter="realtime_start_date<=20130820"' (or ... <=2013-08-20) Maybe accompanied with a new option to declare which columns should be treated as holding ISO-date data: '--isodates=realtime_start_date,realtime_end_date' And finally, a wishlist for 'join' options. These missing features can be worked around, but it would be easier if they existed: 1) '--filter' with two conditions instead of just one; in the realtime context the natural application is to do a "within" check: realtime_start_date <= somedate <= realtime_end_date 2) '--aggr="seq:last"', to get the last matching item, where the number of that last item can vary between different key-matched sequences. Thanks, Sven