It's difficult to select only one, but I would dare to choose this idea.
Advanced Raster Cell Value Calculation
Although FME supports raster manipulations widely, I think there still is a big room to improve efficiency especially for calculating cell values. The RasterCellCoercer + some transformers or script + NumericRasterizer could cover most use cases, but it would be too complicated and too slow! I hope this idea will be realized in the near future.
Not the idea I was going to pitch, but I can't find the one I was looking for, so I see merit in your suggestion.
My first choice was going to be the focal analysis pitch I gave on stage at the FME User conference, but since I can't track down that idea, my second choice is a variation of Dave's
WorkspaceRunner - Identify when all child processes are complete
Currently the workspace runner allows us to run the child workspaces one at a time, waits for them to complete and reports back on their success or failure, OR launches up to 7 workspaces at a time, but doesn't track the process.
So there is no way to take advantage of simultaneous processing and once all the child processes are complete continue the workflow.
A case i encounter many times is dealing with datasets that are too large to process the entire dataset in one shot. The solution is to tile the dataset in the parent workspace and then process each tile via a workspaceRunner, and then reassemble all the tiles for any further processing. However, in this scenario we cannot take advantage of the concurrent processing, because if we are not waiting for the job to complete, the parent workspace is trying to reassemble the tiles before they are finished being processed.
I see two possible ways to improve this:
1) on the WorkspaceRunner don't make Wait for Job to Complete: Yes and Maximum Number of Concurrent Workspaces mutually exclusive. Allow up to 7 concurrent workspaces, but each feature is released only when it's process is complete.
2) the trigger features exist the WorkspaceRunner with the id of the child process as an attribute. There is a new transformer that checks whether a process has been complete (via the id attribute).
My preference is for number 1.
Yes, we saw that and are definitely going to implement it as soon as we can.
It's difficult to select only one, but I would dare to choose this idea.
Advanced Raster Cell Value Calculation
Although FME supports raster manipulations widely, I think there still is a big room to improve efficiency especially for calculating cell values. The RasterCellCoercer + some transformers or script + NumericRasterizer could cover most use cases, but it would be too complicated and too slow! I hope this idea will be realized in the near future.
Yes, I really like that idea. It's something I've long wanted to do, and there was a project I did recently that really needed it. It has my vote!
My first choice was going to be the focal analysis pitch I gave on stage at the FME User conference, but since I can't track down that idea, my second choice is a variation of Dave's
WorkspaceRunner - Identify when all child processes are complete
Currently the workspace runner allows us to run the child workspaces one at a time, waits for them to complete and reports back on their success or failure, OR launches up to 7 workspaces at a time, but doesn't track the process.
So there is no way to take advantage of simultaneous processing and once all the child processes are complete continue the workflow.
A case i encounter many times is dealing with datasets that are too large to process the entire dataset in one shot. The solution is to tile the dataset in the parent workspace and then process each tile via a workspaceRunner, and then reassemble all the tiles for any further processing. However, in this scenario we cannot take advantage of the concurrent processing, because if we are not waiting for the job to complete, the parent workspace is trying to reassemble the tiles before they are finished being processed.
I see two possible ways to improve this:
1) on the WorkspaceRunner don't make Wait for Job to Complete: Yes and Maximum Number of Concurrent Workspaces mutually exclusive. Allow up to 7 concurrent workspaces, but each feature is released only when it's process is complete.
2) the trigger features exist the WorkspaceRunner with the id of the child process as an attribute. There is a new transformer that checks whether a process has been complete (via the id attribute).
My preference is for number 1.
From a technical point of view I have no idea how we would do that. Thankfully that's not my job! As you say, the crux of the issue is not being able to use simultaneous processing while waiting for the jobs to finish. I agree it's an important issue and I hope you pick up some votes for this.
Hi Mark
As said by others, it is really difficult to choose only one, but I'd really like to see multiple output ports on the PythonCaller et al., c.f. this posting: https://knowledge.safe.com/content/idea/24903/add-rejected-port-to-the-pythoncaller.html
David
Hi Mark
As said by others, it is really difficult to choose only one, but I'd really like to see multiple output ports on the PythonCaller et al., c.f. this posting: https://knowledge.safe.com/content/idea/24903/add-rejected-port-to-the-pythoncaller.html
David
I don't use Python that often, but I can see how that would be useful. Right now I guess you'd need to set an attribute and follow up with a Tester.
I don't use Python that often, but I can see how that would be useful. Right now I guess you'd need to set an attribute and follow up with a Tester.
That's right. First the PythonCaller where you set and expose the attribute, then a Tester/TestFilter, then an AttributeRemover to clean up. It's feels rather clunky.
Actually, what you can do is search for a generic term (like "FME") then use the options on the right hand side to filter by ideas and by a specific username. You do have to create a search first, but FME is likely to be in most ideas - or just search for a single letter like "a" or "e". Hope that helps.
Actually, what you can do is search for a generic term (like "FME") then use the options on the right hand side to filter by ideas and by a specific username. You do have to create a search first, but FME is likely to be in most ideas - or just search for a single letter like "a" or "e". Hope that helps.
That didn't quite work. The search seems to pick up whole words only. Searching for 'e' picks up 'e-mail' and '[A B C D E F]' but not 'the'.
Searching by username with an ideas filter pulls up all the ideas authored and commented on. The username filter does not alter those results.
