On Mon, 2 Jan 2017, Jan Tille wrote: The problem is the --send-data option. You use that to send your dataset to R under the name "gretldata", but you are probably running your script with no dataset open, so gretl gets confused. If you either (a) open an empty datased via "nulldata 10" or something similar or (b) get rid of --send-data in the "foreign" command, everything should be ok. ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------

Jack, Thank you for your answer. Actually, my first try was to use the procedure described in chapters 36.3 and 36.4 and I explicitly sent the data from a Gretl dataset with data. I tried both, non-interactive as well as interactive sessions. I also used R's summary function to check whether the data and the matrix was passed to R. I got the same error message ("Datenfehler" <-> data error (?)) with both procedures. Best, Jan -----Ursprüngliche Nachricht----- Von: Plus.line MailSystem [mailto:cyrus@mailer.plusline.de] Im Auftrag von Riccardo (Jack) Lucchetti Gesendet: Montag, 2. Januar 2017 11:46 An: Gretl list <gretl-users(a)lists.wfu.edu&gt; Betreff: Re: [Gretl-users] Problems when trying to interac with R On Mon, 2 Jan 2017, Jan Tille wrote: The problem is the --send-data option. You use that to send your dataset to R under the name "gretldata", but you are probably running your script with no dataset open, so gretl gets confused. If you either (a) open an empty datased via "nulldata 10" or something similar or (b) get rid of --send-data in the "foreign" command, everything should be ok. ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------

Am 02.01.2017 um 16:07 schrieb Jan Tille: I was unable to replicate the problem. Could you post a self-contained example please? (Your code is incomplete for us because we don't know what the matrix "funds" is initially. I created it myself but as I said, I don't get the error.) BTW, you mentioned a different R version that you have. You just have to change gretl's settings (paths in the preferences) to use that other installation, I think. No need to have admin rights for that. Could you also try a newer gretl version than 2016b? thanks, sven

Am 02.01.2017 um 18:13 schrieb Jan Tille: OK, I see, it's the missings. More precisely: The matrix saved by R (via gretl.export) represents the missings as "NaN", while gretl seems to prefer "nan" (lowercase). (also @Allin:) I manually changed all "NaN" to "nan" in the saved file and then gretl's mread() accepts it. Jan, as a workaround for your case until this is fixed, could you somehow restrict your data to have no missings already in gretl, before you call R, then it should work. hth, sven

On Tue, 3 Jan 2017, Riccardo (Jack) Lucchetti wrote: I agree with Jack's diagnosis here: I think this problem is (mostly) already fixed in git. The following script works OK: <hansl> matrix M = mnormal(3, 2) M[1,2] = 0/0 # introduce a NaN value print M mwrite(M, "M.mat", 1) foreign language=R M1 <- gretl.loadmat("M.mat") gretl.export(M1, "M2") end foreign matrix M2 = mread("M2.mat", 1) print M2 </hansl> The retrieved matrix M2 is the same as the original M. In addition, this modified version also works alright: <hansl> open data4-1 matrix M = mnormal(3, 2) M[1,2] = 0/0 # introduce a NaN value print M mwrite(M, "M.mat", 1) foreign language=R --send-data M1 <- gretl.loadmat("M.mat") gretl.export(M1, "M2") end foreign matrix M2 = mread("M2.mat", 1) print M2 </hansl> Now, if you take the second version and delete the line "open data4-1" (so that there's no dataset to send in response to the --send-data option), then executing the script produces: <output> ... ? end foreign write_data_for_R: failed with err = 15 </output> This is admittedly too cryptic, but it's a "correct" error message in that the function write_data_for_R() has failed, for the reason that there's nothing to write. Allin

Thank you for your answers and apologies for the delayed response, I had to contact our admin, because of a safety feature under Windows, that required admin rights for the first launch of Gretl 2016e git. I have tried to run Allin's first example, but still got an error message ("Datenfehler" <-> "Data error") I have provided the script's printout below (in German only) I have no clue, why the script works on your systems but not on mine. Nonetheless, thank you all. Best, Jan gretl-Version 2016e-git Aktuelle Sitzung: 2017-01-03 15:11 #current session ? matrix M = mnormal(3, 2) ? M[1,2] = 0/0 # introduce a NaN value Warnung: erzeugte nichtendliche Werte #Warning message that nan's were created ? print M M (3 x 2) 1,4152 nan 0,47459 0,078130 0,34524 0,32018 ? mwrite(M, "M.mat", 1) ? foreign language=R ? M1 <- gretl.loadmat("M.mat") ? gretl.export(M1, "M2") ? end foreign wrote matrix C:/Users/JTille/AppData/Roaming/gretl/M2.mat ? matrix M2 = mread("M2.mat", 1) Datenfehler # Error message Fehler bei Skriptausführung: Stopp #message that the script stopped -----Ursprüngliche Nachricht----- Von: Plus.line MailSystem [mailto:cyrus@mailer.plusline.de] Im Auftrag von Allin Cottrell Gesendet: Dienstag, 3. Januar 2017 01:45 An: r.lucchetti(a)univpm.it; Gretl list <gretl-users(a)lists.wfu.edu&gt; Betreff: Re: [Gretl-users] Problems when trying to interac with R On Tue, 3 Jan 2017, Riccardo (Jack) Lucchetti wrote: I agree with Jack's diagnosis here: I think this problem is (mostly) already fixed in git. The following script works OK: <hansl> matrix M = mnormal(3, 2) M[1,2] = 0/0 # introduce a NaN value print M mwrite(M, "M.mat", 1) foreign language=R M1 <- gretl.loadmat("M.mat") gretl.export(M1, "M2") end foreign matrix M2 = mread("M2.mat", 1) print M2 </hansl> The retrieved matrix M2 is the same as the original M. In addition, this modified version also works alright: <hansl> open data4-1 matrix M = mnormal(3, 2) M[1,2] = 0/0 # introduce a NaN value print M mwrite(M, "M.mat", 1) foreign language=R --send-data M1 <- gretl.loadmat("M.mat") gretl.export(M1, "M2") end foreign matrix M2 = mread("M2.mat", 1) print M2 </hansl> Now, if you take the second version and delete the line "open data4-1" (so that there's no dataset to send in response to the --send-data option), then executing the script produces: <output> ... ? end foreign write_data_for_R: failed with err = 15 </output> This is admittedly too cryptic, but it's a "correct" error message in that the function write_data_for_R() has failed, for the reason that there's nothing to write. Allin _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users

Am 03.01.2017 um 16:15 schrieb Allin Cottrell: Yes, by coincidence I was already using the brand new snapshot an hour ago or so, and Jan's script now ran fine. Jan, if you don't want to unnerve your admin again, in general it might be advisable to keep the missings out of the stuff that is passed to R in the first place (if that's doable). My feeling is that it would be cleaner anyway, apart from avoiding this particular bug. cheers, sven

I've to admit that I haven't followed the whole discussion. But, Jan, you could replace the NANs by some arbitrary value (e.g. '0' or -999), pass the matrix to gretl and impose the NANs there again by: <hansl> # Check elementwise if matrix x.=0; if zero replace by NA, # otherwise replace by the corresponding value of matrix m matrix m = zeros(1,2) matrix m = (m ? m : NA) </hansl Artur Am 03.01.2017 um 17:40 schrieb Jan Tille: