On Mon, 30 Jun 2014, Sven Schreiber wrote:

> Am 29.06.2014 20:22, schrieb Allin Cottrell:
>> On Sat, 28 Jun 2014, Sven Schreiber wrote:
>>
>>> I keep running into situations where gretl doesn't recognize the dates
>>> in my csv files. Now instead of asking for more heuristics and
>>> cleverness, I'm wondering whether it would be possible for the user to
>>> specify the format pattern at the time of opening/importing the file.
>>> E.g. let the user specify the string "yyyy~mm" to denote monthly data
>>> with strange separation characters (this is just an example, not my
>>> actual case).
>>>
>>> I don't know whether in the file-open dialogs there is actually a place
>>> to do that, though....
>>
>> For special cases one has the "join" command with its --tkey option.
>> I suppose --tkey could be added for plain "open", although as you
>> imply it wouldn't be trivial to find a suitable place for this in
>> the GUI.
>>
>> But could you tell us what sort of CSV dates are not being
>> recognized? Unless they're wildly idiosyncratic I think our
>> heuristics could probably be modified to handle them.
>>
>
> See attached. It's monthly data with date format "01/mm/yyyy", and I
> guess the redundant leading "01/" confuses gretl.

Thanks, Sven. Gretl was trying to interpret the dates as MM/DD/YYYY. That 
interpretation "works", but produces something that's not a sensible time 
series (data for the first 12 days of each year) and therefore gets 
rejected.

So now in CVS if we don't get a sensible interpretation we try looping 
back and using DD/MM/YYYY.

Allin


Reply via email to