We have been working on a v4 system upgrade here and I had the same question setting up the schema for the unique constraint (max length in particular)
Looking at the AIDL data we had the following data shape through 2018 (v3):
Total AIDL Google Purchase Token Count: 476,778
Min Len: 165
Max Len: 187
Avg.Len: 174
While developing for v4 today I snagged some purchase tokens from our license tester purchase activity to do this rather small empirical analysis. Today we had only 15 we could consider and each of those was 144 in length with zero variation. This was a testing license users generating receipt data for us in app using the actual purchase flow (and I suspect that is why they are equal in length.)
the v4 sample:
144 avg/min/max len
![The length of the last 15 captured purchase token string lengths]()
As we get more data (especially from our live early access release coming up) I will endeavor to post more current data (prod me in three months if I don't return. Happy to share findings.) In the AIDL system I had a max text size limit of 400 on the field btw. I think I am just going to keep that and log any occurrence that exceeds that and would require more.
NOTE: I should point out all our data for v4 is consumable, the v3 data was one time entitlements. I don't how much a difference that makes...
NOTE: While working on this v4 system, v5 dropped (2022-05-11)... Looks like its mostly enhancements for subs. I will address those later. We're not enabling subscriptions just yet, but intend too.
edit I initially said v3 everywhere, I meant v4 billing for the new data... our previous system was Billing v3 circa ~2018.
