Am 01.09.2020 um 15:45 schrieb Sven Schreiber: > > without going into detail right now: I try to "join" something and when > I hit OK in the dialog there's no reaction. I'm expecting some kind of > error message. > Trying the same thing in the console just gives a "data error". This was a text file with a string-valued key column (countries) and semicolon as the delimiter. When I changed the semicolons to commas, it works in the console. I guess gretl's heuristics might see the semicolon as part of the string value.

Am 01.09.2020 um 20:50 schrieb Allin Cottrell: > On Tue, 1 Sep 2020, Sven Schreiber wrote: > >> Am 01.09.2020 um 15:45 schrieb Sven Schreiber: >> Trying the same thing in the console just gives a "data error". This was >> a text file with a string-valued key column (countries) and semicolon as >> the delimiter. When I changed the semicolons to commas, it works in the >> console. I guess gretl's heuristics might see the semicolon as part of >> the string value. > > Sven, could you post (offlist if you like) the file in question? > Hm, I cannot reproduce this anymore, so I guess something else was happening as well. What remains true, however, is that I cannot get the join GUI dialog to work (the one that appears after choosing append - advanced/join). Trying to imitate a working join command incarnation just yields no reaction from gretl, and no joining is taking place. The context is an opened geojson file in place and the data to be map-plotted resides in the file to be joined; the matching is done via string-valued series.

Am 01.09.2020 um 23:35 schrieb Allin Cottrell: > On Tue, 1 Sep 2020, Sven Schreiber wrote: > >> Am 01.09.2020 um 20:50 schrieb Allin Cottrell: >>> On Tue, 1 Sep 2020, Sven Schreiber wrote: >>> Please send your files anyway. Will do off-list, later. > > You might try this: open founders.geoson from the geoplot examples > dir. Then do /File/Open data and choose founders.csv in "join" mode. > Select "pop" as the variable to import, string-valued "FID" as the > inner key and string-valued "code" as the outer key. This works fine > for me, does it work for you? No it doesn't. Not sure what you literally mean by "open data in join mode" though. I went to File/Append data and chose advanced/join.

With the settings you described I get the same non-reaction. I'm noticing (and have noticed before) that the entry "code" for the outer key is in a grey font instead of black, as if it were somehow disabled.

I'm just noticing that this is the July-27th snapshot, so pre-2020d if that makes a difference. I can upgrade later, but gotta run now.

Am 02.09.2020 um 13:22 schrieb Allin Cottrell: > On Wed, 2 Sep 2020, Sven Schreiber wrote: >> With the settings you described I get the same non-reaction. I'm >> noticing (and have noticed before) that the entry "code" for the >> outer key is in a grey font instead of black, as if it were somehow >> disabled. > > Hmm, strange. > >> I'm just noticing that this is the July-27th snapshot, so pre-2020d >> if that makes a difference. I can upgrade later, but gotta run now. > > Yes, please try with 2020d or current snapshot. > Same thing with 2020d.

We were not double-quoting the filename from which data were to be joined, and the backslashes in Windows filenames were choking the parser. That's now fixed in git and snapshots.

I guess the join GUI can't have been used much on Windows to date, or we'd have heard about this problem months ago.

We were not double-quoting the filename from which data were to be joined, and the backslashes in Windows filenames were choking the parser. That's now fixed in git and snapshots.

Am 03.09.2020 um 22:54 schrieb Sven Schreiber: > Am 02.09.2020 um 22:17 schrieb Allin Cottrell: >> We were not double-quoting the filename from which data were to be >> joined, and the backslashes in Windows filenames were choking the >> parser. That's now fixed in git and snapshots. >> > It seems there's still a little problem with the join dialog. (In the > snapshot downloaded some hours ago.) Apparently I cannot leave empty the > field "using the name...", although AFAICS that should just > automatically use the outer column label for the series's name. (As > suggested by the grey-color text "same as above".) > > But when I type in a new name it's looking good I think. > My testing shows that the behavior is different in Linux and Windows. On Linux with current git it works fine. On Windows with the latest (yesterday's) snapshot I still get no reaction when the defaults are left as-is. In the text fields I have to manually delete all the "hint" entries like "same as above", "same as inner" etc. to make it work. Perhaps it would be interesting to test on Linux with GTK2, but my understanding is that you (Allin) are using that all the time anyway?