Re: [Gretl-devel] redefining "NA"
On Tue, 21 Aug 2018, Sven Schreiber wrote:
Am 21.08.2018 um 21:32 schrieb Allin Cottrell:
> On Mon, 20 Aug 2018, Riccardo (Jack) Lucchetti wrote:
>
>> Stata has a "saveold" command. As long as _new_ versions of gretl can
save
>> in the "old" format, we should be ok. How about
>>
>> store --oldbinary
>
> OK, I've now taken the plunge: NA is now NaN (internally) in git
> and snapshots. And I've followed Jack's suggestion of adding an
> --oldbinary option to "store" (not yet documented but it will be
> soon).
Sorry, I wanted to reply earlier - doesn't Stata (and other
programs in similar situations) have a format version for their
datafiles? I think it would be good if the conventions and
incarnation of the data file would be visible to the user. OTOH I
guess it's not easy to do because by definition old gretl versions
cannot be updated to recognize this format information? But
perhaps for potential future changes such a meta information tag
should be introduced?
Gretl datafiles (gdt and gdtb) have for many years contained version
information. For example, the included Ramanathan dataset
data4-1.gdt starts (after the XML preamble) with:
<gretldata version="1.3" name="data4-1" frequency="1"
startobs="1"
endobs="14" type="cross-section">
This hasn't been a big deal for the most part, because incompatible
changes to the format have been few and far between. In the present
case we've upped the gretldata version to 1.4, and the signal for
current gretl to expect old-style NAs in a gdtb file is that its
version is less than 1.4.
We could easily expose gretldata version information to users if
there's reason to do so.
Allin