Opinions on the gRPC work by NI on Github?

[drjdpowell]

NI is working on gRPC and LabVIEW.   I am trying to give them feedback on their Github page.  In particular, I think they are going the wrong way in trying to create a Scripting-black-box solution rather than a simpler gRPC Library, and that this is a major danger to the project.  gRPC looks like it could be important, and we have a stake in making this project successful. 

If anyone else would like to comment on this, please see https://github.com/ni/grpc-labview/issues/208

[Neil Pate]

James, I found the NI automatically generated stuff quite hard to understand. I spent time stipping it down to the VIs with the black banners (under those are just the DLL calls) and actually it really is not complicated, its just packed up in a way that is completely not usable for me.

[drjdpowell]

18 minutes ago, Neil Pate said:

James, I found the NI automatically generated stuff quite hard to understand. I spent time stripping it down to the VIs with the black banners (under those are just the DLL calls) and actually it really is not complicated, its just packed up in a way that is completely not usable for me.

I did the exact same thing!  Posted my version here: https://github.com/ni/grpc-labview/issues/206 

I made a small helper class to allow easier configuration of each Procedure with a name and two message clusters:

[ShaunR]

On 12/9/2022 at 11:07 AM, drjdpowell said:

I think they are going the wrong way in trying to create a Scripting-black-box solution rather than a simpler gRPC Library

Well it seems to basically be NI's "Protocol Buffers" - scripted compilation to create [de]sterilizers. I lost interest at that point.

[infinitenothing]

Protocol buffers is one part and I think a cross platform scripted method of defining a typedef is cool.

I definitely got lost a couple times figuring out what was "my code" and what was the toolkit and I had to reinstall the toolkit at least once.

I think if I were making the tool, for a unary server, I would have used VI references similar to "nonlinear curve fit" uses them instead of an event structure. The event structure comes with too much baggage and doesn't enforce a response.

I still don't fully understand how to manage a streamed response. Shouldn't the read timeout be set at each place the data is read instead of a global? Seems like it lends itself better to a queue or something.