ForestDB update_seq on view query less than db's update_seq

[sethrosetter]

---

• Version: 1.3.0-49
• Client OS: Mac OSX
• Server: N/A

They query seems to be skipping the design doc?

To reproduce:

• Launch 1.3.0-49 LiteServ
• Create a db
• Add 10 docs (not sure if number has any effect)
• Add a view

http POST http://localhost:59851/db

with body

{"_id":"_design/test","views":{"basic":{"map":"function(doc) { emit(doc._id, doc.foo) }"}}}

• Query the view

http://localhost:59851/cbl_views/_design/test/_view/basic?update_seq=true

Expected:

{"update_seq":11,"rows":[{"key":"cbl_views_0","value":"bar","id":"cbl_views_0"},{"key":"cbl_views_1","value":"bar","id":"cbl_views_1"},{"key":"cbl_views_2","value":"bar","id":"cbl_views_2"},{"key":"cbl_views_3","value":"bar","id":"cbl_views_3"},{"key":"cbl_views_4","value":"bar","id":"cbl_views_4"},{"key":"cbl_views_5","value":"bar","id":"cbl_views_5"},{"key":"cbl_views_6","value":"bar","id":"cbl_views_6"},{"key":"cbl_views_7","value":"bar","id":"cbl_views_7"},{"key":"cbl_views_8","value":"bar","id":"cbl_views_8"},{"key":"cbl_views_9","value":"bar","id":"cbl_views_9"}],"offset":0,"total_rows":10}

Actual:

{"update_seq":10,"rows":[{"key":"cbl_views_0","value":"bar","id":"cbl_views_0"},{"key":"cbl_views_1","value":"bar","id":"cbl_views_1"},{"key":"cbl_views_2","value":"bar","id":"cbl_views_2"},{"key":"cbl_views_3","value":"bar","id":"cbl_views_3"},{"key":"cbl_views_4","value":"bar","id":"cbl_views_4"},{"key":"cbl_views_5","value":"bar","id":"cbl_views_5"},{"key":"cbl_views_6","value":"bar","id":"cbl_views_6"},{"key":"cbl_views_7","value":"bar","id":"cbl_views_7"},{"key":"cbl_views_8","value":"bar","id":"cbl_views_8"},{"key":"cbl_views_9","value":"bar","id":"cbl_views_9"}],"offset":0,"total_rows":10}

Doing a GET on the /db, I see an update_seq of 11

[snej]

This may be a performance problem because if the view's update_seq (lastSequenceIndexed) is less than the db's lastSequence, it will be indexed. So this would lead to a no-op index operation on every query.

From a quick look into CBForest, this seems to be due to a bug in the MapReduceIndexWriter class. It should update the view's lastSequenceIndexed to the last sequence of the db enumeration, not just the last sequence that the view actually ran through the map function.