问题
I'm trying to query a MongoDB via the R driver rmongodb. The following query works on the cmd line (result: 204,915):
db.col1.count(
{
$or: [
{'status.time':{$gt: ISODate('2013-09-10 00:00:00')}},
{'editings.time':{$gt: ISODate('2013-09-10 00:00:00')}}
]
} );
Translating this into R, I tried:
d<-strptime('2013-09-10', format='%Y-%m-%d')
buf <- mongo.bson.buffer.create()
mongo.bson.buffer.start.array(buf, "$or")
mongo.bson.buffer.start.object(buf, 'status.time')
mongo.bson.buffer.append(buf, "$gt", d)
mongo.bson.buffer.finish.object(buf)
mongo.bson.buffer.start.object(buf, 'editings.time')
mongo.bson.buffer.append(buf, "$gt", d)
mongo.bson.buffer.finish.object(buf)
EDIT: This is what the query prints in R:
>mongo.bson.from.buffer(buf)
$or : 4
status.time : 3
$gt : 9 79497984
editings.time : 3
$gt : 9 79497984
Executing the query using...
mongo.count(mongo, db1.col1, query=mongo.bson.from.buffer(buf))
...gives me "-1". I tried several variants of the BSON, all with the same result. Using only one of the conditions (without the $or array) works, however. Does anyone see my mistake?
BTW: I'm aware of the thread rmongodb: using $or in query, however, the suggested answer to use the alternative driver RMongo does not satisfy other requirements of my code.
回答1:
your way of creating an mongo bson array is wrong. You are missing the parts
mongo.bson.buffer.start.object(buf, "0")
...
mongo.bson.buffer.finish.object(buf)
mongo.bson.buffer.start.object(buf, "1")
...
mongo.bson.buffer.finish.object(buf)
For a working example please check the latest comment on: https://github.com/mongosoup/rmongodb/issues/17
I hope this works for now. I am working on an easier solution!
回答2:
To avoid having to compose the sequence of mongo.bson.buffer-statements I wrote a package (rmongodbHelper) that will translate a JSON or a list() to a BSON object which can then be used with rmongodb.
First let's setup the environment:
library(rmongodb)
# install rmongodbHelper package from GitHub
library(devtools)
devtools::install_github("joyofdata/rmongodbHelper")
library(rmongodbHelper)
# the MongoDB instance
ns <- "dbx.collx"
M <- mongo.create()
mongo.is.connected(M)
mongo.remove(M, ns, json_to_bson("{}"))
# inserting a number of dummy objects
# JSON keys currently are expected to be wrapped in double quotes!
objs <- c(
'{"_id":"__int(0)", "dates":{}}',
'{"_id":"__int(1)", "dates":{"a":"__time(2013-01-01)", "b":"__time(2013-01-01)"}}',
'{"_id":"__int(2)", "dates":{"a":"__time(2013-01-01)", "b":"__time(2014-01-01)"}}',
'{"_id":"__int(3)", "dates":{"a":"__time(2014-01-01)", "b":"__time(2013-01-01)"}}',
'{"_id":"__int(4)", "dates":{"a":"__time(2014-01-01)", "b":"__time(2014-01-01)"}}'
)
for(obj in objs) {
mongo.insert(M, ns, json_to_bson(obj))
}
Let's see via MongoDB shell if they were successfully inserted:
> use dbx
switched to db dbx
> db.collx.find().pretty()
{ "_id" : 0, "dates" : { } }
{
"_id" : 1,
"dates" : {
"a" : ISODate("2013-01-01T00:00:00Z"),
"b" : ISODate("2013-01-01T00:00:00Z")
}
}
[...]
{
"_id" : 4,
"dates" : {
"a" : ISODate("2014-01-01T00:00:00Z"),
"b" : ISODate("2014-01-01T00:00:00Z")
}
}
Now let's search for documents with a query:
# searching for those objects
# JSON keys currently are expected to be wrapped in double quotes!
json_qry <-
'{
"$or": [
{"dates.a":{"$gt": "__time(2013-06-10)"}},
{"dates.b":{"$gt": "__time(2013-06-10)"}}
]
}'
cur <- mongo.find(M, "dbx.collx", json_to_bson(json_qry))
while(mongo.cursor.next(cur)) {
print(mongo.cursor.value(cur))
}
And this is what we get in the end:
_id : 16 2
dates : 3
a : 9 -211265536
b : 9 1259963392
_id : 16 3
dates : 3
a : 9 1259963392
b : 9 -211265536
_id : 16 4
dates : 3
a : 9 1259963392
b : 9 1259963392
- keys - also operators like $or - need to be put in double quotes.
"x":3
will lead to 3 being casted as double"x":"__int(3)"
will lead to 3 being casted as integer
来源:https://stackoverflow.com/questions/19952705/using-or-array-in-query