Question

I’ve been trying to use the aggregate function to group date fields by year:

db.identities.aggregate([
{
    $group : {
        _id : { year : {$year : "$birth_date"}},
        total : {$sum : 1}
        }
    }   
])

Some of my dates however fall before 1970 and being a Windows user I get a nasty error about gmtime:

{
    "errmsg" : "exception: gmtime failed - your system doesn't support dates before 1970",
    "code" : 16422,
    "ok" : 0
}

I know the obvious answer now is for me to get a virtual machine running or something but I was just curious if there were any work-arounds for windows (Windows 7 in my case). Failing that how much of a performance hit would storing the date as a nested object be i.e:

birth_date : {
  year : 1980,
  month : 12,
  day : 9
}

I’m not too sure how hectic that would be with indexes etc.

Any advice appreciated!

Answer 1

Some versions of Windows have been known to work. By any chance, are you using a 32-bit OS? The code in question is here, and depends upon the gmtime_s() implementation.

If this collection is simply for aggregation queries, you can certainly get by with storing date components in an object. I’d suggest abbreviating the field names (e.g. y, m, d) to save on storage, since the field strings are present in each stored document. The trade-off here is that none of the aggregation date operators can be used. You may want to store the timestamp as a signed integer (e.g. ts) so that you can easily do range queries if necessary.