I recreated your test workbench in 2017.1 (17539), and the output was as expected. Was the StatisticsCalculator placed within 17529 or 17539? Maybe there was an update to the transformer?
I had a similar issue with the StatisticsCalculator a while ago. Re-installing the software did the trick for me. Have you tried running your workspace on a different machine with the same FME Build?
Hi @reservoirdog. Also, I would suggest connecting a Logger to the StatisticsCalculator to see whether anything is recorded in the log.
Hi @reservoirdog
could you please share your translation log?
Have you tried running this workspace on build 17259 once again? Did you get the expected results?
@reservoirdog
could you please also check if Emailer or ChartGenerator work on your b.17539 or b.17650? StatiscticsCalculator, Emailer, and ChartGenerator are Python-based transformers. If something is wrong with Python on your machine/in your FME install, they all will fail... I hope.
Sadly it worked for me just fine as well (I just ran your workspace on a variety of builds, all produced correct output). This definitely smells like something is wrong with your python installation. I'd hope our log file might shed light on it, but I'd double check what Python you're set to use.
Thanks for all your inputs...
I have this on two different machines (Win Server 2008R2 and 2012R2) both "coming" from
Build 17259. On our productive machine I reverted back to 17291 (last 2017.0.1), and statistics work well as normal.
@LenaAtSafe
Logfile attached:
statistics-problem.txtHave you tried running this workspace on build 17259 once again? Did you get the expected results?
-> Yes, works fine again...
Seems I should try a clean re-install... well get back to you
@reservoirdog
could you please also check if Emailer or ChartGenerator work on your b.17539 or b.17650? StatiscticsCalculator, Emailer, and ChartGenerator are Python-based transformers. If something is wrong with Python on your machine/in your FME install, they all will fail... I hope.
ChartGenerator works well, tested in the same workspace...
I had a similar issue with the StatisticsCalculator a while ago. Re-installing the software did the trick for me. Have you tried running your workspace on a different machine with the same FME Build?
Yes, I have this problem on two different machines...
Both coming from 17259... will have to try a clean re-install...
I recreated your test workbench in 2017.1 (17539), and the output was as expected. Was the StatisticsCalculator placed within 17529 or 17539? Maybe there was an update to the transformer?
Thx for your input. I tried with Builds 17529 as well as the newest 17650.
StatisticsCalculator works well (again) on 17259 and 17291
Hi @reservoirdog
could you please share your translation log?
Have you tried running this workspace on build 17259 once again? Did you get the expected results?
Thx, Translation bug is attached to latest reply.
And yes, everything works well back on 17259 and 17291
Sadly it worked for me just fine as well (I just ran your workspace on a variety of builds, all produced correct output). This definitely smells like something is wrong with your python installation. I'd hope our log file might shed light on it, but I'd double check what Python you're set to use.
Thx for your inputs, see my latest reply for the attached log file...
Did a clean re-install following https://knowledge.safe.com/articles/962/re-installing-fme-if-there-is-a-problem-with-the-i.html with success.
StatisticsCalculator back working with latest build 17650.
lesson learned: Don't do in-place updates anymore...
Thx for your support.
Did a clean re-install following https://knowledge.safe.com/articles/962/re-installing-fme-if-there-is-a-problem-with-the-i.html with success.
StatisticsCalculator back working with latest build 17650.
lesson learned: Don't do in-place updates anymore...
Thx for your support.
@reservoirdog, glad this worked for you!
We have seen several clients with this issue. Upgrading FME to a newer build (2017+) causes the StatisticsCalculator to produce erroneous results. Unfortunately, the problem is not consistently reproducable (although I have managed to reproduce it on occasion). A simple re-install of FME is often not sufficient to fix the issue. Usually the affected computer needs to have all traces of FME removed before re-installing.
It would be good if Safe could do some more investigation into this issue as it's embarrasing to have to ask clients to uninstall and re-install and hope that fixes the problem...
We have seen several clients with this issue. Upgrading FME to a newer build (2017+) causes the StatisticsCalculator to produce erroneous results. Unfortunately, the problem is not consistently reproducable (although I have managed to reproduce it on occasion). A simple re-install of FME is often not sufficient to fix the issue. Usually the affected computer needs to have all traces of FME removed before re-installing.
It would be good if Safe could do some more investigation into this issue as it's embarrasing to have to ask clients to uninstall and re-install and hope that fixes the problem...
I don't know if this is related, but I found a bug with earlier versions of 2017 where the stats calculator would drop the 0 (ie if it was 10 it would now be classified as 1). It got resolved in 17652 (2017.1.1.1).
We have seen several clients with this issue. Upgrading FME to a newer build (2017+) causes the StatisticsCalculator to produce erroneous results. Unfortunately, the problem is not consistently reproducable (although I have managed to reproduce it on occasion). A simple re-install of FME is often not sufficient to fix the issue. Usually the affected computer needs to have all traces of FME removed before re-installing.
It would be good if Safe could do some more investigation into this issue as it's embarrasing to have to ask clients to uninstall and re-install and hope that fixes the problem...
Thanks for pointing this out @nic_ran -- we've taken a peek and are doing a bit of additional digging. It is true that the StatisticsCalculator makes use of our Python subsystem in a different way than most of our other Python stuff, and as such, we have some suspicions. We agree that this is not a good situation.
Thanks for pointing this out @nic_ran -- we've taken a peek and are doing a bit of additional digging. It is true that the StatisticsCalculator makes use of our Python subsystem in a different way than most of our other Python stuff, and as such, we have some suspicions. We agree that this is not a good situation.
Thanks @daleatsafe. I know this is going to be a tricky one to track down but it would be really nice to get it resolved.
Thanks @daleatsafe. I know this is going to be a tricky one to track down but it would be really nice to get it resolved.
So we've been able to duplicate what people are seeing BUT only by manually using an older .pyc file in a newer install. It is as though the new installer failed to erase the old implementation. This would imply someone upgrading should uninstall first, then clean their directory. We will keep digging to see how this situation could ever occur.