I asked this at mongo-usr, mongo-dev and Stack Overflow with no luck. I may be obtuse and am missing something fundamentally simple.

Given this data with a unique compound index:

db.employees.drop()
db.employees.insert( { employeenum : 1, check : "A" } )
db.employees.insert( { employeenum : 1, check : "B" } )
db.employees.insert( { employeenum : 2, check : "A" } )
db.employees.insert( { employeenum : 2, check : "B" } )
db.employees.insert( { employeenum : 2, check : "C" } )
db.employees.insert( { employeenum : 5, check : "E" } )
db.employees.insert( { employeenum : 6, check : "A" } )
db.employees.ensureIndex( { employeenum: 1, check : 1 }, {unique: true} )

If I want the next document in the index after {{

}}. I can do this:

db.employees.find({ query: { $or: [ { employeenum: { $gt: 5 } }, { check: { $gt: "E" } } ] }, $min: { employeenum: 5, check: "E" }, $maxScan: 2 })

which returns {{

}}.

But how do I traverse backwards from {{

}}? How do I fetch {{

}}? The way I am doing it:

{ query: { $or: [ { employeenum: { $lt: 5 } }, { check: { $lt: "E" } } ] }, $hint: { employeenum: -1, check: -1 }, $min: { employeenum: 5, check: "E" }, $maxScan: 2 }

requires a reverse index, a very inefficient solution. Is there a better way? B-tree indexes are bi-directional.

One suggested:

db.Emp.find({$or:[{employeenum{$lt:5}}]}).sort({employeenum:-1,check:-1}).limit(1)

And yes that will gets me 2-C. Now, how to go from 2-C to 2-B? This:

db.employees.find({$or:[{employeenum:{$lt:2}},{check:{$lt:"C"}}]}).sort({employ??eenum:-1,check:-1}).limit(1)

returns {{

}} I need a general way to traverse the index from a point on the index. I am using the C++ driver but using the shell here for illustration. We are developing a rapid development tool using mongoDB as the underlying database. This is a (presumably) simple use case: A window displays a single document from a collection with a compound index. The user presses a button to display the next document using the compound index. Or another button to display the previous.

I did try:

db.employees.find({ query: { $or: [ { employeenum: { $lt: 5 } }, { check: { $lt: "E" } } ] }, $max: { employeenum: 5, check: "E" }, orderby: { employeenum: -1, check: -1 }, $maxScan: 2,  $explain: false })

But that crashed the server every time which raises a concern that a shell users could crash the server with an (apparently) malformed query in a mission critical application such as ours.