Am 04.11.2024 um 17:28 schrieb klaus.hasenbach(a)web.de:

> but surprisingly the bundl request does not work with all data, for > instance not for > > provider = "OECD" > database = "DSD_​IMTS@DF_​IMTS" > bundle spec = defbundle("mask","DNK.W.X.C.M.XDC.Y.N") > bs = dbnomics_get_multiple(provider, database, 1000, 0, spec) OK I see, and I agree it's surprising. I checked what is returned from the relevant gretl-dbnomics internals, and I'm seeing: ? =req.output {"_meta":{"args":{"align_periods":false,"dataset_code":"DSD_\u00e2\u0080\u008bIMTS@DF_\u00e2\u0080\u008bIMTS","dimensions":{},"facets":false,"format":"json","limit":1000,"metadata":false,"observations":true,"offset":0,"provider_code":"OECD","q":"","series_code":"DNK.W.X.C.M.XDC.Y.N"},"version":"22.1.17"},"message":"Dataset 'OECD\/DSD_\u00e2\\x80\\x8bIMTS@DF_\u00e2\\x80\\x8bIMTS' not found"} So I believe on the dbnomics server side they are inserting weird unicode characters (the \u00e2\u0080\u008b above), rendering the resulting json invalid. I have no idea where that comes from or why, my current guess would simply be it's a bug over there.

* Given 4, it seems that the strange bytes must have actually been in Klaus's script (and not added somehow via an email quirk). So the question then is how they got into Klaus's script in the first place. My guess is by copy-and-paste from somewhere.

Am 05.11.2024 um 03:30 schrieb Allin Cottrell: > A few more thoughts on the messed-up dbnomics database identifier. > > Here's what we know: > > 1) The second scripting example that Klaus posted contained strange > "invisible" bytes in the supposed database identifer, as it appeared > on the mailing list. > > 2) If you try to get data from dbnomics using the identifier as shown, > it doesn't work. > > 3) If you delete the strange bytes, leaving plain underscores, you get > the data from dbnomics OK. A follow-up on this: I had another real use case, which involved copying the codes from the db.nomics website. I got an error but remembered Klaus's problem and so I deleted the copied underscores and re-typed them in gretl's editor. Still an error - the cause was that deleting the underscores didn't affect the invisible characters right *after* the underscores. So that was quite subtle. Bottom line: I wonder whether gretl's editor could use some gtksourceview library internal switch to display invisible bytes by default in the future. Those characters do not belong in hansl scripting files (I'd say), and so displaying them in the form of some weird unicode codepoint would immediately show the problem.

Am 25.11.2024 um 22:03 schrieb Cottrell, Allin: > In current gretl git and snapshots, we automatically delete any > zero-width (and so invisible) UTF-8 characters in text pasted from the > clipboard into the script editor, the console, and the dialog box > where you specify a particular dbnomics series to access. I think that > should cover all cases where web-junk might cause a problem. Thanks, Allin, sounds very good. (But haven't tested yet.)

On Tue, Nov 26, 2024 at 11:32 AM Sven Schreiber <sven.schreiber(a)fu-berlin.de&gt; wrote: > > I cannot test the console because pasting from the system clipboard > doesn't work there -- this is an unrelated problem which I have > mentioned before, I think. > > Then I checked the small dialog window that I get under File / Databases > / Dbnomics / Specific series: Pasting fails there, too - both with > ctrl-v keyboard shortcut or via the context menu (although the paste > entry is active/not greyed out). Pasting into the console and the dbnomics series entry box are now enabled in the Windows snapshots.