FeedIterator unexpectedly pre-fetches next page

Describe the bug The FeedIterator<T> fetches a page of results before it is requested. A single call of ReadNextAsync() causes two POSTs to the cosmos API, rather one.

To Reproduce Here is a sample application that illustrates the problem. Note ReadNextAsync() is only called once, but the console logs two API calls.

Expected behavior FeedIterator should only fetch pages upon request.

Actual behavior FeedIterator seems to be calling for a second page of results on its own. If this is by design, then I think it is flawed, as it makes a huge assumption that we care to get more than the number of items requested.

Environment summary SDK Version: 3.4.1 OS Version: Windows 10

Additional context Here is a sample of the console’s output. Note that two POSTs occur, despite only having called ReadNextAsync() once. Also note that the contents of the second POST show the next set of results (items 5-9).

POST dbs/FeedTest/colls/Test
{"_rid":"dIVbALcqJUk=","Documents":[
  {"r":{"id":"3cc75122-20d1-4d14-ba41-c1f328d2e9f3","Index":0,"Partition":"test","_rid":"dIVbALcqJUkBAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkBAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff2d1b101d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"015aa151-203a-4b85-9c7b-bd99ade327e6","Index":1,"Partition":"test","_rid":"dIVbALcqJUkCAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkCAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff4ecf101d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"b74259ec-ce80-4881-931c-39355e95e23a","Index":2,"Partition":"test","_rid":"dIVbALcqJUkDAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkDAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff6345a01d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"b2d53634-923c-4a54-a9f2-09ff3a4e1c97","Index":3,"Partition":"test","_rid":"dIVbALcqJUkEAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkEAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff699df01d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"9555e4d7-e75f-47e2-83d3-84c1b85757bf","Index":4,"Partition":"test","_rid":"dIVbALcqJUkFAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkFAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff6dd4901d5\"","_attachments":"attachments\/","_ts":1573619207}}
],"_count":5}

POST dbs/FeedTest/colls/Test
{"_rid":"dIVbALcqJUk=","Documents":[
  {"r":{"id":"1b51b134-ef3c-4f61-b42b-486391bc2e7e","Index":5,"Partition":"test","_rid":"dIVbALcqJUkGAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkGAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff7320701d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"a4d7a76d-683d-46ec-94c1-c820a3572fdf","Index":6,"Partition":"test","_rid":"dIVbALcqJUkHAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkHAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff77f7201d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"189306b0-aae2-4b4f-9ff8-e0b566943fe4","Index":7,"Partition":"test","_rid":"dIVbALcqJUkIAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkIAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff7cf4e01d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"348cf465-de43-420e-9c30-11116dc4a755","Index":8,"Partition":"test","_rid":"dIVbALcqJUkJAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkJAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff81b2101d5\"","_attachments":"attachments\/","_ts":1573619207}},
  {"r":{"id":"297d734e-ba44-409b-8c0c-95c44aa13983","Index":9,"Partition":"test","_rid":"dIVbALcqJUkKAAAAAAAAAA==","_self":"dbs\/dIVbAA==\/colls\/dIVbALcqJUk=\/docs\/dIVbALcqJUkKAAAAAAAAAA==\/","_etag":"\"00000000-0000-0000-99da-8ff8684101d5\"","_attachments":"attachments\/","_ts":1573619207}}
],"_count":5}

[j82w]

No need to apologize. I should have been more clear that we had functional bug that is taking priority, and that this will get prioritized next. I’m just trying to be as transparent as possible with the reasoning behind the prioritization of fixes. We really appreciate the detailed bug report with the sample. We’ll try to get the issue fixed as soon as we can.

[bartelink]

eta: 3.7.0-preview2 does not fix the issue, and neither does the custom build in the other issue

🎉 3.9.0-preview fixes my problem - reducing RU consumption back to levels as per V2 (using preview-3 now which is also healthy).

Any rough eta on when we’ll see a 4.x preview with this fix included?