Distinct results while saving/validating pipeline vs executing pipeline

[darshthakkar]

Hi Team,

Lately, I observed distinct results getting generated during compile time (i.e. saving a pipeline) vs run time (i.e. executing a pipeline). Has anyone observed this behavior?

For instance, consider a simple example of reading a flat file, doing some transformations, using joins and exporting it as a csv/excel file. When we save the pipeline and the Snap Execution has been set to Validate & Execute, a flat file will be generated in Manager with default 50 records (this may be ranging from 1-2000 depending on each user’s preview count settings). The file generated during saving a pipeline is different than one generated after executing a pipeline.

What I observed was, the joins didn’t work during compile time whereas they worked during run time.

My use case doesn’t need to export a file but to update new records in Salesforce, what’s happening over here is that, during compile time, a vital field data is going blank and while running the pipeline, the data is going as is. The concern here is "As I'm considering brand new records ONLY ", compile time execution ingests wrong data and run time wouldn’t be able to ingest the correct data as the record was already ingested (new fields only, run time wouldn’t fetch that data as it’s no longer a new record)

Any help/thoughts on this would be highly appreciated.
Thanks!

Best Regards,
Darsh

[ptaylor]

Remember that a Join will only join the data that’s available from its input views – the output of the upstream snaps attached to its inputs. Those data sets are also constrained by the Preview Document Count if you’re validating. So a Join that might normally find many matches when the complete data is available during an execution might find no matches when it’s only dealing with a 50 record subset from each input view. I think this might not be obvious. It’s easy to get the misimpression that the Preview Document Count is only a limit on the amount of data that we’re displaying but all the rest of the data is still processed. That’s not the case. When a snap reaches that count on its output, it actually stops and doesn’t output any more data, even invisibly, even if there’s a lot more data than that available. This is important to understand.

Could that explain what you’re seeing?

[darshthakkar]

It does a bit however I specified before that during validating/saving a pipeline, say for instance, ID = 100 will display Phone Number as “Null” and during pipeline execution, it will display Phone Number as 123456.

As I had access to upstream systems, I knew ID=100 has a phone number 123456 (this is achieved via Join in snapLogic) but I don’t get this while validating my pipeline. If ID=100 wasn’t available in the output preview, I would have been fine with that considering it didn’t come in the first 50 records however it shows ID=100 and Phone Number = “Null”; this is concerning for me!

Moreover, as the Join didn’t work as expected, ID=100 inserted a new record in Salesforce (as ID=100 was a new record) with Phone number = “Null” (when the pipeline was validated as Salesforce Upsert had its snap execution set to Validate & Execute) and this is a weird behavior in my honest opinion. The same has been observed while saving a pipeline if snap execution is Validate & Execute.

I wouldn’t have realized this as I wasn’t checking the output preview but when I validated the records in Salesforce, I was seeing a lot of “Null” values, that made me go to the upstream system, randomly pick up 10 IDs and check their values. To my surprise, those IDs had some data and still were going as Null. After investigating, I came to a conclusion that minute changes in the pipeline and then saving it is causing the issue, thus I quickly disabled the salesforce upsert snap, made all the relevant changes, changed snap execution of Salesforce Upsert to Execute ONLY from Validate & Execute. I still disable the Salesforce Upsert whenever the requirement changes and I have to modify my pipeline as I feel that Joins are not functioning during validating/saving pipeline but they do during executing pipeline (which is a bit absurd)

[ptaylor]

So in the output of the Join, the Phone Number you’re expecting is null. Do you see the correct value of this Phone Number in one of the inputs to the Join? The preview will show you all the input data so you be able to see what should be correctly joined. Can you see specific input records that should have been joined but weren’t?

[darshthakkar]

Yes, that’s correct. The phone number I’m expecting for an ID (based on the data in the upstream systems) comes as NULL. This behavior is intermittent, at times, output preview of Join will fetch the expected data, other times it won’t. As there are multiple joins on my pipeline, the final mapper where I rename the cols to ingest into Salesforce doesn’t include the fetched data from Join even though Output preview of Join displayed something.

Yes, I do see the correct value of Phone Number in one of the inputs to the Join.

Yes, I do see specific input records that should have been joined but outputting Null values. Reiterating this, this happens in saving/validating pipeline NOT with pipeline execution. With pipeline execution, it works as expected.

As the phone number value from Joins are coming out as Null during saving/validating pipeline, it ingests (inputs) those to the next snap i.e. Salesforce Upsert (as the snap execution was set to Validate & Execute)

[ptaylor]

Can you please show us some screenshots? I’d like to see screenshots showing:

• The pipeline.

• The configuration settings of the Join.

• The Pipeline Validation Statistics, particularly the Join.

• The Pipeline Execution Statistics, particularly the Join.

• Ideally, some preview data for the inputs and output of the Join in a case where it didn’t work correctly, at least showing a particular pair of records that should have been joined correctly and the output record where you’re seeing nulls instead of the values from the input records.

Also, please Download the actual data from these previews and save the json. There may be a subtle difference in the data type of the values for the field that you’re joining on that may become more evident if we can see the JSON (e.g. maybe it’s the string “100” in one and the number 100 in the other).