I use MATCHES() AQL function to search for entries in the arango database matching an example. This feature works nice for plain examples, but I cannot get it work properly with the nested features. See example:
RETURN MATCHES(
{ "a" : { "c" : 1 }, "b" : 1 },
{ "a" : { "c" : 1 } },
false
)
This returns true, however if I try:
RETURN MATCHES(
{ "a" : { "c" : 1, "b" : 1 }},
{ "a" : { "c" : 1 } },
false
)
It returns false !! (I expected to return true)
I have read that this is known in the Query by example section
https://www.arangodb.com/docs/stable/data-modeling-documents-document-methods.html#query-by-example
Their solution is to use dot notation, but it does not work in AQL
Following their example:
RETURN MATCHES(
{ "a" : { "c" : 1, "b" : 1 } },
{ "a.c" : 1 },
false
)
returns false (and I would expect to return true)
How can I then, use the MATCHES() for nested attributes?
FYI: I use arangodb v3.5.5-1
Clarification:
I want to get a match of { "a" : { "c" : 1, "b" : 1 } } by giving { "a" : { "c" : 1 } } as example
I've posted the Issue in ArangoDB repository: https://github.com/arangodb/arangodb/issues/12541
MATCHES compares attributes - it doesn't care what type the attributes have, so if you are trying to match nested objects, they have to have the same attributes/values; it follows that you can not have arbitrarily deep structures overlayed and checked for correspondence.
In the given example you can pick out the substructure with LET and use MATCHES against that
LET doc = { "a" : { "c" : 1, "b" : 1 }}
LET a = doc.a
RETURN MATCHES(
a,
{ "c": 1}
)
To leverage the arangojs capability to use paths to peek into the structure, you can write a user function that uses query by example and call that from AQL (https://www.arangodb.com/docs/stable/aql/extending.html).
Nota Bene: The arangodb client library of your language should provide convenient access to registering user functions (e.g. class AqlUserFunctions in arangodb-php).
Related
Here is the problem I am facing:
I am using arangob 3.7 and arangojs driver.
I have following collections:
collection A { _key, data }
collection B { _key, aDataList[A._key] }
I have tried the following
FOR bdoc IN B
FILTER bdoc._key == "some_key"
FOR adoc IN A
FILTER adoc._key IN bdoc.aDataList[*]
RETURN MERGE(bdoc, adoc)
This query returns the objects which falls in to the criteria specified.
But the problem I am facing is the bdoc.aDataList[] order is not same as the one in the actual B document collection.
Lets say here is the sample list:
bdoc.aDataList[ 1, 2, 3 ]
How it need to be updated?
bdoc.aDataList[
{
"_key" : 1,
"data" : "somedata"
},
{
"_key" : 2,
"data" : "somedata"
},
{
"_key" : 3,
"data" : "somedata"
}
]
How to properly replace the aDataList[A.Key] with aDataList[A] values using a single aql query?
Any help would be appreciated
I have found an answer :)
FOR bDoc IN B
FILTER bDoc.key == “somekey”
LET finalData = ( FOR bDocItem IN bDoc.aDataList
FOR aDoc IN A
FILTER bDocItem[“_key”] == aDoc._key
RETURN aDoc)
RETURN { "_key" : bDoc.key, aDataList: finalData }
Instead of traversing keys of A , I traverse through the array. Thus order is preserved
I'm fairly new to mongo and while I can manage to do most basic operations with the $in, $or, $all, ect I can't make what I want to work.
I'll basically put a simple form of my problem. Part of my documents are list of number, eg :
{_id:1,list:[1,4,3,2]}
{_id:2,list:[1]}
{_id:3,list:[1,3,4,6]}
I want a query that given a list(lets call it L), would return me every document where their entire list is in L
for example with the given list L = [1,2,3,4,5] I want document with _id 1 and 2 to be returned. 3 musn't be returned since 6 isn't in L.
"$in" doesn't work because it would also return _id 3 and "$all" doesn't work either because it would only return _id 1.
I then thought of "$where" but I can't seem to find how to bound an external variable to the js code. What I call by that is that for example :
var L = [1,2,3,4,5];
db.collections('myCollection').find({$where:function(l){
// return something with the list "l" there
}.bind(null,list)})
I tried to bind list to the function as showed up there but to no avail ...
I'd glady appreciate any hint concerning this issue, thanks.
There's a related question Check if every element in array matches condition with an answer with a nice approach for this scenario. It refers to an array of embedded documents but can be adapted for your scenario like this:
db.list.find({
"list" : { $not : { $elemMatch : { $nin : [1,2,3,4,5] } } },
"list.0" : { $exists: true }
})
ie. the list must not have any element that is not in [1,2,3,4,5] and the list must exist with at least 1 element (assuming that's also a requirement).
You could try using the aggregation framework for this where you can make use of the set operators to achieve this, in particular you would need the $setIsSubset operator which returns true if all elements of the first set appear in the second set, including when the first set equals the second set; i.e. not a strict subset.
For example:
var L = [1,2,3,4,5];
db.collections('myCollection').aggregate([
{
"$project": {
"list": 1,
"isSubsetofL": {
"$setIsSubset": [ "$list", L ]
}
}
},
{
"$match": {
"isSubsetofL": true
}
}
])
Result:
/* 0 */
{
"result" : [
{
"_id" : 1,
"list" : [
1,
4,
3,
2
],
"isSubsetofL" : true
},
{
"_id" : 2,
"list" : [
1
],
"isSubsetofL" : true
}
],
"ok" : 1
}
So I'm using Node.js with MongoDB for my web application. I'm having some trouble creating a text index for my schema and searching for text within an array. I've looked at the mongo docs but haven't found anything related to this specifically.
My current implementation searches successfully on regular String values, but querying for text matching in [String]'s don't return anything.
Here's my REST call:
...console.log("Query string: " + str);
var qry = {
"$text": {
"$search": str
}
};
model.find(qry, function (err, results) {...
And when I create my schema:
var blah = new Schema({
foo : String,
bar : [String],
...
blah.index({
foo: 'text',
bar: 'text'
});
Any query won't return the results that match in bar. A query string for something within foo works fine.
Double check that you've created the correct indexes on the correct collections and the queries are being issued to the correct collections. Indexing an array works for me:
> db.test.drop()
> db.test.insert({ "_id" : 0, "a" : "dogs are good" })
> db.test.insert({ "_id" : 1, "a" : "I like dogs", "b" : ["where's my dog?", "here, have a cat"] })
> db.test.insert({ "_id" : 2, "b" : ["she borrowed my dog", "my frogs are croaking"] })
> db.test.ensureIndex({ "a" : "text", "b" : "text" })
> db.test.find({ "$text" : { "$search" : "dogs" } }, { "_id" : 1 })
{ "_id" : 0 }
{ "_id" : 2 }
{ "_id" : 1 }
Okay, I finally figured it out! Turns out, grunt serve doesn't update indexes in the database. I had created a text index for "foo" only and that didn't update when I added "bar" to the index. I had to run - in mongo shell:
db.dropDatabase()
The next time I ran it, the database was recreated and the proper indexes were set. If anyone else runs across this issue, try running db.getIndexes().
I have an array and it contains duplicate values in BOTH the ID's, is there a way to remove one of the duplicate array item?
userName: "abc",
_id: 10239201141,
rounds:
[{
"roundId": "foo",
"money": "123
},// Keep one of these
{// Keep one of these
"roundId": "foo",
"money": "123
},
{
"roundId": "foo",
"money": "321 // Not a duplicate.
}]
I'd like to remove one of the first two, and keep the third because the id and money are not duplicated in the array.
Thank you in advance!
Edit I found:
db.users.ensureIndex({'rounds.roundId':1, 'rounds.money':1}, {unique:true, dropDups:true})
This doesn't help me. Can someone help me? I spent hours trying to figure this out.
The thing is, I ran my node.js website on two machines so it was pushing the same data twice. Knowing this, the duplicate data should be 1 index away. I made a simple for loop that can detect if there is duplicate data in my situation, how could I implement this with mongodb so it removes an array object AT that array index?
for (var i in data){
var tempRounds = data[i]['rounds'];
for (var ii in data[i]['rounds']){
var currentArrayItem = data[i]['rounds'][ii - 1];
if (tempRounds[ii - 1]) {
if (currentArrayItem.roundId == tempRounds[ii - 1].roundId && currentArrayItem.money == tempRounds[ii - 1].money) {
console.log("Found a match");
}
}
}
}
Use an aggregation framework to compute a deduplicated version of each document:
db.test.aggregate([
{ "$unwind" : "$stats" },
{ "$group" : { "_id" : "$_id", "stats" : { "$addToSet" : "$stats" } } }, // use $first to add in other document fields here
{ "$out" : "some_other_collection_name" }
])
Use $out to put the results in another collection, since aggregation cannot update documents. You can use db.collection.renameCollection with dropTarget to replace the old collection with the new deduplicated one. Be sure you're doing the right thing before you scrap the old data, though.
Warnings:
1: This does not preserve the order of elements in the stats array. If you need to preserve order, you will have retrieve each document from the database, manually deduplicate the array client-side, then update the document in the database.
2: The following two objects won't be considered duplicates of each other:
{ "id" : "foo", "price" : 123 }
{ "price" : 123, "id" : foo" }
If you think you have mixed key orders, use a $project to enforce a key order between the $unwind stage and the $group stage:
{ "$project" : { "stats" : { "id_" : "$stats.id", "price_" : "$stats.price" } } }
Make sure to change id -> id_ and price -> price_ in the rest of the pipeline and rename them back to id and price at the end, or rename them in another $project after the swap. I discovered that, if you do not give different names to the fields in the project, it doesn't reorder them, even though key order is meaningful in an object in MongoDB:
> db.test.drop()
> db.test.insert({ "a" : { "x" : 1, "y" : 2 } })
> db.test.aggregate([
{ "$project" : { "_id" : 0, "a" : { "y" : "$a.y", "x" : "$a.x" } } }
])
{ "a" : { "x" : 1, "y" : 2 } }
> db.test.aggregate([
{ "$project" : { "_id" : 0, "a" : { "y_" : "$a.y", "x_" : "$a.x" } } }
])
{ "a" : { "y_" : 2, "x_" : 1 } }
Since the key order is meaningful, I'd consider this a bug, but it's easy to work around.
consider the following records:
user record
{
"_id" : ObjectId("5234ccb7687ea597eabee677"),
"class" : [
{ "_id" : ObjectId("5234ccb7687ea597eabee671", "num" : 10, "color" : "blue" },
{ "_id" : ObjectId("5234ccb7687ea597eabee672", "num" : 100, "color" : "blue" }
]
}
this user has two class sub records, now I need a query that finds all users that have class property where "class._id" has a value of at least one users "class._id"
here is a more detail example:
suppose there is four user:
A:{_id:432645624232345,class:[{_id:123,name:'foo'}]}
B:{_id:432645624232555,class:[{_id:555,name:'foo'},{_id:123,name:'foo'}]}
C:{_id:432645344232345,class:[{_id:555,name:'foo'},{_id:111,name:'www'}]}
D:{_id:432644444232345,class:[{_id:222,name:'sss'},{_id:555,name:'www'},{_id:123,name:'foo'}]}
now if B login , I need to query all the user whose class subdocument contains at least one document which's _id==555 or _id==123 (555 and 123 come from B user), in this case the query result should be:
A:{_id:432645624232345,class:[{_id:123,name:'foo'}]} // match _id=123
B:{_id:432645624232555,class:[{_id:555,name:'foo'},{_id:123,name:'foo'}]} //match _id=123 and _id=555
C:{_id:432645344232345,class:[{_id:555,name:'foo'},{_id:111,name:'www'}]} //match _id=555
D:{_id:432644444232345,class:[{_id:222,name:'sss'},{_id:555,name:'www'},{_id:123,name:'foo'}]} ///match _id=123 and _id=555
which is all the user.
so far i get this:
{"class._id" : { $in : ["5234ccb7687ea597eabee671", "5234ccb7687ea597eabee672"] } }
but when different user login the class._id query condition is different. So is there any operator to do this
{"class._id" : { $in : req.user.class } }
hope I made myself clear.
In order to achieve what you want, first you must isolate the class _ids in an array, and then use it in the query argument.
var classIds = [];
var i = 0;
while (i < req.user.class.length) {
classIds.push(req.user.class[i]._id);
i++;
}
After that you can use classIds array in the query:
{"class._id" : { $in : classIds } }
The following query condition would give you all the users that have at least one class with id equal to any of the elements in the given array:
{"class._id" : { $in : ["5234ccb7687ea597eabee671", "5234ccb7687ea597eabee672"] } }
In the array for the $in clause you may provide any id's you needed , comma separated.
In addition, if you needed such, the below query condition should check for existence of nested document within "class" property that has a property "_id" :
{ "class._id" : { $exists : true } }
Both conditions should work no matter if "class._id" is a single-valued property or an array (mongo supports that).