Analyze Test Step with a large number of reports
Answered
We are looking to analyze some measurements for YTD.
We setup the filters to allow 200,000 test reports, and it appears that we have ~150k for the part number I am looking to analyze.
The Test Step and Analysis is able to generate the graphs, but I do have to click the "Wait or Exit" dialog in chrome at least once, but not a big deal.
If I try to export the data to Excel though, that appears to crash the webpage.
I would like to get the raw data out, preferably in the WATs_StepDetails Excel format, but if there is another way, that would be great too.
Do I need to use the WATs API and filter by the Operation/Part# etc, and locate the specific measurement, and basically loop through all the reports to build the excel report myself?
Or is there an API to get the Excel-style report from the server programmatically (ie not through the web interface)?
Considering that we can test millions of a single part number per year, limiting the data to 1k or 10k test reports, we would be missing a lot of the data.
We do understand that Excel is not the best tool for that large of a dataset, but just using it as an example, since it is an option to select on the Test Step Detail page.
-
Official comment
Hi Kenny.
You may use the Export wizard to export your data to excel. You can filter down your results and select the step of interest, much like you would do in the Test step yield and analysis.
Select "UUT reports" as the type, "Selected steps" as datasource and "Excel" as format.
There are some limitations to the amount of reports you can export at once though, so you may need to run the export more than once to export all the data of interest.
Comment actions -
Hi Kenny,
may I add that we also provide the measurements as aggregated data, better suited to review large datasets.
This will let you analyze the min, max, and average of the data down to one hour intervals. You can find this using the Process Capability Analysis report, and clicking the "View details (aggregated view)" for each test step. The aggregated data can also be accessed using our REST API and will let you review millions of samples in one query.Can I also ask; what is the use case and what are you trying to achieve/read from the data?
-
We are tracking some specific metrics related to response time from our internal database, and saving the values as measurements. The response time directly affects the amount of products we can produce daily.
Every test cycle will save the response times as a measurement, so we can utilize the Cpk, avg, etc metrics.
We have been working with the team in charge of the database to reduce the times. As part of the analysis of how the slow response is affecting our production, we need to put a cost to it, so we are using the YTD metrics.
Since we have 450k measurements YTD for some of the measurements, the graphing can take almost 1 min. If I wanted to share the "source data" via the Excel report, I have been unable to Export the data using the Excel button for a data set this large.
It appears that maybe 10-15k is about the limit, before the browser locks up waiting for the report to be generated. -
Thanks, that makes sense.
I would think the aggregated measurement data could be used for this, but unfortunately there is currently no option to export the aggregated measurement data from the GUI.
What you can do, is to use the REST API endpoint: api/app/AggregatedMeasurements to load these directly to excel. Using a filter like this (replace with your part number, testOp code)
{
"partNumber": "277288",
"testOperation": "50",
"dateFrom": "2023-01-01T00:00:00Z",
"dateGrouping": 1
}This will give you a dataset grouped into days (with dateGrouping:1) and every group will have a count and an average value. Simply multiply the average by the count for each group and summarize to get the total values.
The date grouping can be set to give groups pr. hour, day, week, quarter and year. Selecting year would give you one group with a total count and an average measurement.
Please sign in to leave a comment.
Comments
4 comments