Re: [Gretl-devel] join filter and >=
Am 09.09.2013 22:45, schrieb Allin Cottrell:
On Mon, 9 Sep 2013, Sven Schreiber wrote:
> Ok, so I have stripped this down to the attached short script with the
> sample data in the also attached csv file. I'm still getting the wrong
> results; have tested with existing and new workfiles, no change. I
> somehow think I must be blind to not spot the obvious mistake, but I
> just don't see it.
>
> This is now with the latest snapshot on Windows.
Could there be some weird difference on Windows? (In case there's any
out-of-syncness with the snapshots I just did a Windows build and upload.)
I'm running (on Linux) a slightly modified version of your script (just
adds INDPRO too, attached) on your data, and here's the non-null portion
of the constructed data:
INDPRO startdates
2001:01 102.7055 20090327
2001:02 102.1096 20090327
2001:03 101.7946 20090327
2001:04 101.5296 20090327
2001:05 100.7886 20090327
2001:06 100.1489 20090327
2001:07 99.7063 20090327
2001:08 99.3386 20090327
2001:09 98.9988 20090327
2001:10 98.4253 20090327
2001:11 97.9427 20090327
2001:12 97.9025 20090327
Allin
and here's my output when I run your script:
2001:01 25 91.0867 20100625
2001:02 26 102.1096 20090327
2001:03 27 90.2978 20100625
2001:04 28 101.5296 20090327
2001:05 29 89.4668 20100625
2001:06 30 100.1489 20090327
2001:07 31 99.7063 20090327
2001:08 32 88.2057 20100625
2001:09 33 98.9988 20090327
2001:10 34 98.4253 20090327
2001:11 35 87.0363 20100625
2001:12 36 97.9025 20090327
The errors are consistent between the two columns, it's the wrong lines
that get picked sometimes.
This is not your ultra-recent build, since I'm on a slow connection
right now.
The mystery continues...
-s