Entity Framework Core Performance Optimization
Entity Framework Core Performance Optimization
Last year I talked about Entity Framework Core. It is a easy and feature rich ORM which makes working with database in a .NET environment typesafe. But even though it makes things easy, there are ambiguous cases which can take us off guard. Today we will see four of this cases and how to deal with them.
- Client evaluation
- Iteration
- Include and ThenInclude
- NoTracking
For the following examples, I will be using SQLite with Entity Framework Core.
tl;dr
- Make sure that the query constructed in c# uses function that can be translated to SQL,
- Make sure that there isn’t an abnormal amount of queries created and that it does not iter item per item,
- Make sure to use
Include
andThenInclude
for object relation to include them after query execution, before query execution it is not needed, - Use
NoTracking
for readonly queries to disable tracking on entity to yield better performance.
1. Client evaluation
The following example illustrates the first case, client evaluation
:
var query = _dbContext.Posts
.Where(p => p.BlogId == blogId && p.AuthorId == authorId && p.Title.Contains(title, StringComparison.OrdinalIgnoreCase));
return await query.ToListAsync();
We are executing a query and using .Contains(string, StringComparison.OrdinalIgnoreCase)
, the compiler allows us to write it as it is valid code and Entity Framework runs properly. But if we look at the logs, we’ll see a warning:
warn: Microsoft.EntityFrameworkCore.Query[20500]
The LINQ expression 'where [p].Title.Contains(__title_2, OrdinalIgnoreCase)' could not be translated and will be evaluated locally.
This indicates that the where
clause will be executed locally therefore all elements that satisfy p.BlogId == blogId && p.AuthorId == authorId
will be fetched locally then filtered locally, in log we can see the query:
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (7ms) [Parameters=[@__blogId_0='?', @__authorId_1='?'], CommandType='Text', CommandTimeout='30']
SELECT "p"."Id", "p"."AuthorId", "p"."BlogId", "p"."Content", "p"."Title"
FROM "Posts" AS "p"
WHERE ("p"."BlogId" = @__blogId_0) AND ("p"."AuthorId" = @__authorId_1)
This could be problematic if there were a lot of data as it would degrade the performance. Changing it to .Contains(string)
:
var query = _dbContext.Posts
.Where(p => p.BlogId == blogId && p.AuthorId == authorId && p.Title.Contains(title));
return await query.ToListAsync();
We can see from the query that it uses instr(...)
which is supported by SQLite:
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (14ms) [Parameters=[@__blogId_0='?', @__authorId_1='?', @__title_2='?' (Size = 9)], CommandType='Text', CommandTimeout='30']
SELECT "p"."Id", "p"."AuthorId", "p"."BlogId", "p"."Content", "p"."Title"
FROM "Posts" AS "p"
WHERE (("p"."BlogId" = @__blogId_0) AND ("p"."AuthorId" = @__authorId_1)) AND ((instr("p"."Title", @__title_2) > 0) OR (@__title_2 = ''))
2. Iterations
The second case is about iterations
, consider the following:
var result = (await _dbContext.Blogs
.Select(b => new
{
url = b.Url,
posts = b.Posts.Select(p => new { title = p.Title, author = p.Author.Name })
})
.ToListAsync());
The execution of the query happens at .ToListAsync()
, but in the query, we have an anonymous type which within the construct maps to another anonymous type. The problem with this query is that the query can’t be translated to a SQL query. Therefore the result is a query on Blogs
then one query per blog to get the posts:
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (15ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
PRAGMA foreign_keys=ON;
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (4ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
SELECT "b"."Url", "b"."Id"
FROM "Blogs" AS "b"
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (0ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
PRAGMA foreign_keys=ON;
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (6ms) [Parameters=[@_outer_Id='?'], CommandType='Text', CommandTimeout='30']
SELECT "p"."Title", "p.Author"."Name" AS "author"
FROM "Posts" AS "p"
INNER JOIN "Authors" AS "p.Author" ON "p"."AuthorId" = "p.Author"."Id"
WHERE @_outer_Id = "p"."BlogId"
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (0ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
PRAGMA foreign_keys=ON;
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (1ms) [Parameters=[@_outer_Id='?'], CommandType='Text', CommandTimeout='30']
SELECT "p"."Title", "p.Author"."Name" AS "author"
FROM "Posts" AS "p"
INNER JOIN "Authors" AS "p.Author" ON "p"."AuthorId" = "p.Author"."Id"
WHERE @_outer_Id = "p"."BlogId"
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (0ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
PRAGMA foreign_keys=ON;
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (1ms) [Parameters=[@_outer_Id='?'], CommandType='Text', CommandTimeout='30']
SELECT "p"."Title", "p.Author"."Name" AS "author"
FROM "Posts" AS "p"
INNER JOIN "Authors" AS "p.Author" ON "p"."AuthorId" = "p.Author"."Id"
WHERE @_outer_Id = "p"."BlogId"
Again this wasn’t expected as knowing SQL, we would have expected a single query with some sort of JOIN
. To do that, we can force the execution and translation of the query earlier:
var result = (await _dbContext.Blogs
.Include(b => b.Posts)
.ThenInclude(p => p.Author)
.ToListAsync())
.Select(b => new
{
url = b.Url,
posts = b.Posts.Select(p => new { title = p.Title, author = p.Author.Name })
});
Which results in an expected query:
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (0ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
PRAGMA foreign_keys=ON;
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (0ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
SELECT "b"."Id", "b"."Url"
FROM "Blogs" AS "b"
ORDER BY "b"."Id"
info: Microsoft.EntityFrameworkCore.Database.Command[20101]
Executed DbCommand (0ms) [Parameters=[], CommandType='Text', CommandTimeout='30']
SELECT "b.Posts"."Id", "b.Posts"."AuthorId", "b.Posts"."BlogId", "b.Posts"."Content", "b.Posts"."Title", "p.Author"."Id", "p.Author"."Name"
FROM "Posts" AS "b.Posts"
INNER JOIN "Authors" AS "p.Author" ON "b.Posts"."AuthorId" = "p.Author"."Id"
INNER JOIN (
SELECT "b0"."Id"
FROM "Blogs" AS "b0"
) AS "t" ON "b.Posts"."BlogId" = "t"."Id"
ORDER BY "t"."Id"
3. Include and ThenInclude
The next case is regarding relational entities. Here Posts
is a collection within Blogs
. With the following query, we can use Posts
:
var result = (await _dbContext.Blogs
.Select(b => new
{
url = b.Url,
posts = b.Posts.Select(p => new { title = p.Title, author = p.Author.Name }) // <-- Author is accessible
})
.ToListAsync());
But if we move the execution to earlier, Posts
will no longer be available as the collection will not be loaded:
var result = (await _dbContext.Blogs.ToListAsync())
.Select(b => new
{
url = b.Url,
posts = b.Posts.Select(p => new { title = p.Title, author = p.Author.Name }) // <-- System.ArgumentNullException: 'Value cannot be null.' -- 'Posts' is null.
});
That code will yield a null reference as b.Posts
will be null. To load it we need to use .Include()
which allows us to load a collection from Blog
then we can use .ThenInclude()
to load a relation from Post
, here the Author
:
var result = (await _dbContext.Blogs
.Include(b => b.Posts)
.ThenInclude(p => p.Author) // <-- Include Posts in Blog then include Author in Post in the query
.ToListAsync())
.Select(b => new
{
url = b.Url,
posts = b.Posts.Select(p => new { title = p.Title, author = p.Author.Name }) // <-- Posts and Author are included in the query
});
4. NoTracking
The last case is more of a performance improvement. Each query from Entity Framework Core returns object that are tracked by the DbContext. The tracking allows to know what property changed and when the context is saved with dbContext.SaveChanges()
. But when we just want to query to retrieve values and return them for API calls for example, we can disable the tracking with .AsNoTracking()
.
var result = (await _dbContext.Blogs
.AsNoTracking()
.Include(b => b.Posts)
.ThenInclude(p => p.Author)
.ToListAsync())
.Select(b => new
{
url = b.Url,
posts = b.Posts.Select(p => new { title = p.Title, author = p.Author.Name }) // <-- Posts and Author are included in the query
});
Conclusion
Today we saw four cases where Entity Framework Core can behave in an unexpected manner. To conclude, here is a summary of advices:
- Make sure that the query constructed in c# uses function that can be translated to SQL,
- Make sure that there isn’t an abnormal amount of queries created and that it does not iter item per item,
- Make sure to use
Include
andThenInclude
for object relation to include them after query execution, before query execution it is not needed, - Use
NoTracking
for readonly queries to disable tracking on entity to yield better performance.
Hope you liked this post, see you on the next one!
Comments
Post a Comment