Document count query ignores PartitionKey - azure-cosmosdb

I am looking to get the count of all documents in a chosen partition. The following code however will return the count of all documents in the collection and costs 0 RU.
var collectionLink = UriFactory.CreateDocumentCollectionUri(databaseId, collectionId);
string command = "SELECT VALUE COUNT(1) FROM Collection c";
FeedOptions feedOptions = new FeedOptions()
{
PartitionKey = new PartitionKey(BuildPartitionKey(contextName, domainName)),
EnableCrossPartitionQuery = false
};
var count = client.CreateDocumentQuery<int>(collectionLink, command, feedOptions)
.ToList()
.First();
adding a WHERE c.partition = 'blah' clause to the query will work, but costs 3.71 RUs with 11 documents in the collection.
Why would the above code snippet return the Count of the whole Collection and is there a better solution to for getting the count of all documents in a chosen partition?

If the query includes a filter against the partition key, like SELECT
* FROM c WHERE c.city = "Seattle", it is routed to a single partition. If the query does not have a filter on partition key, then it is
executed in all partitions, and results are merged client side.
You could check the logical steps the SDK performs from this official doc when we issue a query to Azure Cosmos DB.
If the query is an aggregation like COUNT, the counts from individual
partitions are summed to produce the overall count.
So when you just use SELECT VALUE COUNT(1) FROM Collection c, it is executed in all partitions and results are merged client side.
If you want to get the count of all documents in a chosen partition, you just add the where c.partition = 'XX' filter.
Hope it helps you.

I believe this is actually a bug since I am having the same problem with the partition key set in both the query and the FeedOptions.
A similar issue has been reported here:
https://github.com/Azure/azure-cosmos-dotnet-v2/issues/543
And Microsoft's response makes it sound like it is an SDK issue that is x64-specific.

Related

DynamoDB how to get items count for a partition keys using .net core?

How can I get items count for a particular partition key using .net core preferably using Object Persistence Interface or Document Interfaces?
Since I do not see any docs any where, currently I get the number of items count by retrieve all the item and get its count, but it is very expensive to do the reads.
What is the best practices for such item count request? Thank you.
dynamodb is mostly a document oriented key-value db; so its not optimized for functionality of the common relation db functions (like item count).
to minimize the data that is transmitted and to improve speed you may want to do the following:
Create Lambda Function that returns Item Count
To avoid transmitting data outside of AWS; which is slow and expensive.
query options
use only keys in your projection-expression,
reducing the data that is transmitted from db
max page-size, reducing number of calls needed
Stream Option
Streams could also be used for keeping counts; e.g. as described in
https://medium.com/signiant-engineering/real-time-aggregation-with-dynamodb-streams-f93547cfb244
https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/bp-gsi-aggregation.html
Related SO Question
Complexity of finding total records count with partition key in nosql dynamodb table?
I just realized that using low level interface in QueryRequest one can set Select = "COUNT" then when calling QueryAsync() orQuery() will return the count only as a integer only. Please refer to code sample below.
private static QueryRequest getStockRecordCountQueryRequest(string tickerSymbol, string prefix)
{
string partitionName = ":v_PartitionKeyName";
string sortKeyPrefix = ":v_sortKeyPrefix";
var request = new QueryRequest
{
TableName = Constants.TableName,
ReturnConsumedCapacity = ReturnConsumedCapacity.TOTAL,
Select = "COUNT",
KeyConditionExpression = $"{Constants.PartitionKeyName} = {partitionName} and begins_with({Constants.SortKeyName},{sortKeyPrefix})",
ExpressionAttributeValues = new Dictionary<string, AttributeValue>
{
{ $"{partitionName}", new AttributeValue {
S = tickerSymbol
}},
{ $"{sortKeyPrefix}", new AttributeValue {
S = prefix
}}
},
// Optional parameter.
ConsistentRead = false,
ExclusiveStartKey = null,
};
return request;
}
but I would like to point out that this still will consumed the same read units as retrieving all the item and get its count by yourself. but since it is only returning the count as an integer, it is a lot more efficient then transmitting the entire items list cross the wire.
I think using DynamoDB Streams in a more proper way to get the counts for large project. It is just a lot more complicated to implement.

Multiple partitions in COSMOS DB collection

1) I have a Cosmos DB collection with about 500k documents and which is Partitioned by a property "SITEID". In the Query Request Options only one partition key value can be passed. In my case I have queries where the SITEID in (1,2,3,4) needs to be executed where SiteID is the partition key.
For example, my SP is as follows:
SELECT * FROM c WHERE c.SITEID IN
("SiteId1","SiteId2","SiteId3","SiteId4","SiteId5")
AND c.STATUS IN ("Status1","Status2","Status3","Status4")
I am Calling the above SP using the below SQL API code.
await client.ExecuteStoredProcedureAsync<string>(UriFactory.CreateStoredProcedureUri("DBName", "CollectionName", "Sample"),new RequestOptions { PartitionKey = new PartitionKey("SiteId1") })
In the above SQL API Code, PartitionKey property only supports a Single value. Where I need to pass several partition values. Is there any other options to do this?
2) "EnableCrossPartitionQuery" property is only availbale in the FeedOptions but not in the Request Options class. Client.ExecuteStoredProcedureAsync only supports the RequestOptions parameter not FeedOptions. Now I need to execute a Stored Procedure at once and across all partitions. Is there any other options to pass EnableCrossPartitionQuery in ExecuteStoredProcedureAsync method.
E.g)
client.CreateDocumentQuery<Doc>(UriFactory.CreateDocumentCollectionUri("DBName", "CollectionName"), "select * from c", new FeedOptions { EnableCrossPartitionQuery = true }).ToList()
await client.ExecuteStoredProcedureAsync<string>(UriFactory.CreateStoredProcedureUri("DBName", "CollectionName", "Sample"),new RequestOptions { PartitionKey = new PartitionKey("WGC") })
Stored procedures can only be executed against a single partition. There is nothing you can do about that.
They are not considered a query that returns a feed but a request that could return a response of any type. That's they they don't used the FeedOptions but rather the RequestOptions.
You can still execute your query as a normal document query and set the EnableCrossPartitionQuery to true. Cosmos should recognise the partition key in the query and should limit the requests to the specific partition key values.
I say should because this answer suggests that this is the case but there are some comments that say otherwise. I would suggest you check your metrics regarding the amount of collection hits.

DynamoDb - .NET Object Persistence Model - LoadAsync does not apply ScanCondition

I am fairly new in this realm and any help is appreciated
I have a table in Dynamodb database named Tenant as below:
"TenantId" is the hash primary key and I have no other keys. And I have a field named "IsDeleted" which is boolean
Table Structure
I am trying to run a query to get the record with specified "TenantId" while it is not deleted ("IsDeleted == 0")
I can get a correct result by running the following code: (returns 0 item)
var filter = new QueryFilter("TenantId", QueryOperator.Equal, "2235ed82-41ec-42b2-bd1c-d94fba2cf9cc");
filter.AddCondition("IsDeleted", QueryOperator.Equal, 0);
var dbTenant = await
_genericRepository.FromQueryAsync(new QueryOperationConfig
{
Filter = filter
}).GetRemainingAsync();
But no luck when I try to get it with following code snippet (It returns the item which is also deleted) (returns 1 item)
var queryFilter = new List<ScanCondition>();
var scanCondition = new ScanCondition("IsDeleted", ScanOperator.Equal, new object[]{0});
queryFilter.Add(scanCondition);
var dbTenant2 = await
_genericRepository.LoadAsync("2235ed82-41ec-42b2-bd1c-d94fba2cf9cc", new DynamoDBOperationConfig
{
QueryFilter = queryFilter,
ConditionalOperator = ConditionalOperatorValues.And
});
Any Idea why ScanCondition has no effect?
Later I also tried this: (throw exception)
var dbTenant2 = await
_genericRepository.QueryAsync("2235ed82-41ec-42b2-bd1c-d94fba2cf9cc", new DynamoDBOperationConfig()
{
QueryFilter = new List<ScanCondition>()
{
new ScanCondition("IsDeleted", ScanOperator.Equal, 0)
}
}).GetRemainingAsync();
It throws with: "Message": "Must have one range key or a GSI index defined for the table Tenants"
Why does it complain about Range key or Index? I'm calling
public AsyncSearch<T> QueryAsync<T>(object hashKeyValue, DynamoDBOperationConfig operationConfig = null);
You simply cant query a table only giving a single primary key (only hash key). Because there is one and only one item for that primary key. The result of the Query would be that still that single item, which is actually Load operation not Query. You can only query if you have composite primary key in this case (Hash (TenantID) and Range Key) or GSI (which doesn't impose key uniqueness therefore accepts duplicate keys on index).
The second code attempts to filter the Load. DynamoDBOperationConfig's QueryFilter has a description ...
// Summary:
// Query filter for the Query operation operation. Evaluates the query results and
// returns only the matching values. If you specify more than one condition, then
// by default all of the conditions must evaluate to true. To match only some conditions,
// set ConditionalOperator to Or. Note: Conditions must be against non-key properties.
So works only with Query operations
Edit: So after reading your comments on this...
I dont think there conditional expressions are for read operations. AWS documents indicates they are for put or update operations. However, not being entirely sure on this since I never needed to do a conditional Load. There is no such thing like CheckIfExists functionality as well in general. You have to read the item and see if it exists. Conditional load will still consume read throughput so your only advantage would be only NOT retrieving it in other words saving the bandwith (which is very negligible for single item).
My suggestion is read it and filter it in your application layer. Dont query for it. However what you can also do is if you very need it you can use TenantId as hashkey and isDeleted for range key. If you do so, you always have to query when you wanna get a tenant. With the query you can set rangeKey(isDeleted) to 0 or 1. This isnt how I would do it. As I said, would just read it and filter it at my application.
Another suggestion thing could be setting a GSI on isDeleted field and writing null when it is 0. This way you can only see that attribute in your table when its only 1. GSI on such attribute is called sparse index. Later if you need to get all the tenants that are deleted (isDeleted=1) you can simply scan that entire index without conditions. When you are writing null when its 0 dynamoDB wont put it in the index at the first place.

DynamoDB Mapper Query Doesn't Respect QueryExpression Limit

Imagine the following function which is querying a GlobalSecondaryIndex and associated Range Key in order to find a limited number of results:
#Override
public List<Statement> getAllStatementsOlderThan(String userId, String startingDate, int limit) {
if(StringUtils.isNullOrEmpty(startingDate)) {
startingDate = UTC.now().toString();
}
LOG.info("Attempting to find all Statements older than ({})", startingDate);
Map<String, AttributeValue> eav = Maps.newHashMap();
eav.put(":userId", new AttributeValue().withS(userId));
eav.put(":receivedDate", new AttributeValue().withS(startingDate));
DynamoDBQueryExpression<Statement> queryExpression = new DynamoDBQueryExpression<Statement>()
.withKeyConditionExpression("userId = :userId and receivedDate < :receivedDate").withExpressionAttributeValues(eav)
.withIndexName("userId-index")
.withConsistentRead(false);
if(limit > 0) {
queryExpression.setLimit(limit);
}
List<Statement> statementResults = mapper.query(Statement.class, queryExpression);
LOG.info("Successfully retrieved ({}) values", statementResults.size());
return statementResults;
}
List<Statement> results = statementRepository.getAllStatementsOlderThan(userId, UTC.now().toString(), 5);
assertThat(results.size()).isEqualTo(5); // NEVER passes
The limit isn't respected whenever I query against the database. I always get back all results that match my search criteria so if I set the startingDate to now then I get every item in the database since they're all older than now.
You should use queryPage function instead of query.
From DynamoDBQueryExpression.setLimit documentation:
Sets the maximum number of items to retrieve in each service request
to DynamoDB.
Note that when calling DynamoDBMapper.query, multiple
requests are made to DynamoDB if needed to retrieve the entire result
set. Setting this will limit the number of items retrieved by each
request, NOT the total number of results that will be retrieved. Use
DynamoDBMapper.queryPage to retrieve a single page of items from
DynamoDB.
As they've rightly answered the setLimit or withLimit functions limit the number of records fetched only in each particular request and internally multiple requests take place to fetch the results.
If you want to limit the number of records fetched in all the requests then you might want to use "Scan".
Example for the same can be found here

How can I COUNT(DISTINCT) in ServiceStack Ormlite?

I'm writing a paged query in ServiceStack's OrmLite, selecting the total records and the ids of records in the page range. Assuming query is some arbitrary SqlExpression selecting a bunch of records:
var idQuery = query.SelectDistinct(r => r.Id);
var ids = Db.Column<int>(idQuery.Skip(request.Skip).Take(request.Take));
var total = Db.Count(idQuery);
OrmLite generates two queries, one for the ids:
SELECT DISTINCT ...
And one for the total:
SELECT COUNT(*)
I'm trying to get OrmLite to generate SELECT COUNT(DISTINCT Id) for the total query, or perform an equivalent. Is this possible?
In previous versions of OrmLite you would need to use Custom SQL:
var count = db.Scalar<long>(idQuery.Select("COUNT(DISTINCT Id)"));
I've just added support for Sql.CountDistinct in this commit which will let you use a Typed API:
var count = db.Scalar<long>(idQuery.Select(x => Sql.CountDistinct(x.Id)));
This change is available from v4.0.61 that's now available on MyGet.

Resources