Unsigned data types not being handled

[ghansham]

Dear Sir

We need to handle unsigned data types properly

edal-java/cdm/src/main/java/uk/ac/rdg/resc/edal/dataset/cdm/CdmGridDataSource.java

Line 457 in 5b96297

switch (arrLocal.getDataType()) {

[guygriffiths]

Only unsigned bytes are allowed by the CF conventions, by using the byte data type and then specifying the valid_min and valid_max values. I've just fixed that and pushed it to develop.

[ghansham]

Thanks a lot.

…

On Tue, May 12, 2020, 15:56 Guy Griffiths ***@***.***> wrote: Only unsigned bytes are allowed by the CF conventions, by using the byte data type and then specifying the valid_min and valid_max values. I've just fixed that and pushed it to develop. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#126 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAYXFJFEDKZNHMOT3YQIAFLRREP3TANCNFSM4M6UJUMQ> .

[ghansham]

Should then such datasets should be rejected at ingestion stage? On Tue, May 12, 2020, 16:50 Ghansham Sangar <ghanshamsangar@gmail.com> wrote:

…

Thanks a lot. On Tue, May 12, 2020, 15:56 Guy Griffiths ***@***.***> wrote: > Only unsigned bytes are allowed by the CF conventions, by using the byte > data type and then specifying the valid_min and valid_max values. I've > just fixed that and pushed it to develop. > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#126 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAYXFJFEDKZNHMOT3YQIAFLRREP3TANCNFSM4M6UJUMQ> > . >

[ghansham]

You can simplify the logic. There is a data type for UBYTE. Check: https://github.com/Unidata/netcdf-java/blob/master/cdm/core/src/main/java/ucar/ma2/DataType.java On Tue, May 12, 2020, 16:51 Ghansham Sangar <ghanshamsangar@gmail.com> wrote:

…

Should then such datasets should be rejected at ingestion stage? On Tue, May 12, 2020, 16:50 Ghansham Sangar ***@***.***> wrote: > Thanks a lot. > > On Tue, May 12, 2020, 15:56 Guy Griffiths ***@***.***> > wrote: > >> Only unsigned bytes are allowed by the CF conventions, by using the byte >> data type and then specifying the valid_min and valid_max values. I've >> just fixed that and pushed it to develop. >> >> — >> You are receiving this because you authored the thread. >> Reply to this email directly, view it on GitHub >> <#126 (comment)>, >> or unsubscribe >> <https://github.com/notifications/unsubscribe-auth/AAYXFJFEDKZNHMOT3YQIAFLRREP3TANCNFSM4M6UJUMQ> >> . >> >

[guygriffiths]

No, I cannot simplify the logic. Like I said:

Only unsigned bytes are allowed by the CF conventions, by using the byte data type and then specifying the valid_min and valid_max values.

The unsigned byte data type is not supported by the CF conventions. See here: http://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/cf-conventions.html#_data_types

[ghansham]

Okay. Agree. I was trying to say putting the case for UBYTE and not for any other unsigned data types. Its okay for me.

…

On Tue, May 12, 2020, 17:14 Guy Griffiths ***@***.***> wrote: No, I cannot simplify the logic. Like I said: Only unsigned bytes are allowed by the CF conventions, *by using the byte data type* and then specifying the valid_min and valid_max values. The unsigned byte data type is not supported by the CF conventions. See here: http://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/cf-conventions.html#_data_types — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#126 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAYXFJEN7KBUO6HXHBAJBZLRREZANANCNFSM4M6UJUMQ> .

[guygriffiths]

It's probably worth including UBYTE, USHORT, and UINT as well, but the byte/valid_min/valid_max behaviour should stay there (despite being a little unintuitive)

[jonblower]

I guess the unsigned stuff in the Unidata Java libraries is there to support other data formats, like BUFR perhaps? From: Guy Griffiths <notifications@github.com> Reply to: Reading-eScience-Centre/edal-java <reply@reply.github.com> Date: Tuesday, 12 May 2020 at 14:08 To: Reading-eScience-Centre/edal-java <edal-java@noreply.github.com> Cc: Subscribed <subscribed@noreply.github.com> Subject: Re: [Reading-eScience-Centre/edal-java] Unsigned data types not being handled (#126) It's probably worth including UBYTE, USHORT, and UINT as well, but the byte/valid_min/valid_max behaviour should stay there (despite being a little unintuitive) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#126 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABQEQ2J426A7XFE3YJBSMLLRRFC47ANCNFSM4M6UJUMQ>.