This is the same exact question as this, but instead for EF5.
Is it possible now?
We have a Users table that has an int PK, but in our other tables that have columns like InsertBy/UpdateBy, the desire is to use value of the LANID varchar column from the Users table, rather than the UserID.
No it is still not possible. FK must target PK in the principal table because EF still doesn't support unique keys (prerequisite for using non-PK columns).
Related
I find this abit confusing. Iam using memsql column store. I try to understand if there is a way to enforce duplications on specific key (e.g eventId). I found some doc regarding Unenforced Unique but I didnt really understand its intention.
The point of unenforced unique keys is as a hint:
An unenforced unique constraint is informational: the query planner may use the unenforced unique constraint as a hint to choose better query plans.
from https://docs.memsql.com/v6.8/concepts/unenforced-unique-constraints/.
Unfortunately MemSQL does not support (enforced) unique constraints on columnstore tables.
MemSQL now supports unique constraint with version 7+ but can be applied to only single column:
https://docs.memsql.com/v7.1/guides/use-memsql/physical-schema-design/creating-a-columnstore-table/creating-a-columnstore-table/
Your columnstore table definition can contain metadata-only unenforced unique keys, single-column hash keys (which may be UNIQUE), and a FULLTEXT key. You cannot define more than one unique key.
one hack to enable UNIQUE constraint on multi columns is to use a computed column consisting of multiple columns appended and then apply UNIQUE on it which will indirectly enforce uniqueness on multiple columns.
example:
CREATE TABLE articles (
id INT UNSIGNED,
year int UNSIGNED,
title VARCHAR(200),
body TEXT,
SHARD KEY(title),
KEY (id) USING CLUSTERED COLUMNSTORE,
KEY (id) USING HASH,
UNIQUE KEY (title) USING HASH,
KEY (year) USING HASH);
I am not good in English!
There is a table in Cassandra 3.5 which all columns of a row don't come at same time. Unique of table is some columns that are unique in a row together, but some of them are null at first. I can not set them the primary key because of null value. I have identify a column with name id and type uuid in Cassandra.
How can I have a unique key with that columns together in Cassandra?
Is my data model true?
How can I solve this problem?
You can't. It's not a relational DB. Use clustering and/or partitioning keys to add an unique constraint.
See this answer
To store unique values, create a separate table having your unique value as a key. Check if it exists by requesting this table before inserting a row. But beware, even doing this, you cannot ensure it will be unique in your final table if you have two concurrent inserts.
Basically, I would recommend using Cassandra as it really is: A data store. And find a way to implement your business logic where it belongs: in your code.
while designing the datamodel in cassandra. I am stuck while designing the below scenario.
Like One API/Webservice can have multiple parameters(input/output). I don't know the parameters count and its column name as well.
How to design its cassandra datamodel. I am aware that supercolumns are not good to use and alternative good solution is using composite keys. But for my scenario I don't have fixed columns names and count that I can specify as composite keys.
Please see the pic below which I want to model
Secondly how to write its create table statement so that I can specify parameter name as column name.
Please let me know if anything is unclear.
Thanks,
Why not use a map?
http://www.datastax.com/documentation/cql/3.1/cql/cql_using/use_map_t.html
create table foo(
name text,
owner text,
version text,
params map<text, text>,
primary key (name, owner, version)
);
If you're one 2.1, you can create secondary indexes on the map keys / values, which caters to more flexibility if needed.
I'm trying to set up my postgres schema in yesod. The plan is to let yesod generate the tables, but I'll be inserting new records into the tables from a lua script in a C++ program. I'd like one of the tables to have a primary key based on a unique Int64 that comes from the C++ environment. Can I get Persistent to treat this column as the table key, but without the automatic id generation?
I guess I could have a regular persistent-style record ID column and an Int64 column too but that seems wasteful and overly complicated. The Int64 will always be unique and that's what I'll use to do lookups on the table records to see if they exist already.
I think this question was asked on the mail list. The short answer is no, the primary key in Persistent is auto-increment, but you can have secondary indexes
Disclaimer: I jumped to C# 2008 recently and SubSonic 3 (3.0.0.4) at the same time. I haven't used Linq for much of anything in the past.
Is there an easy way to use the foreign key display value for sorting, rather than the FK Id (which is numeric)?
I've added a new Find method in my ActiveRecord.tt to help with sorting based on a string field name but after doing some testing I realized that even though its working as it should be, I am not handling foreign key fields at all (they are just sorting by their value).
Even if I need to change how I am accessing the data it is early enough in the project to do that. Just looking for suggestions.
LINQ is your friend in this situation, you just need to join your two objects and then sort by the property from your foreign object:
var primaryObjectsSorted =
from primaryObjects in PrimaryObject.All()
join foreignObjects in ForeignObject.All()
on primaryObjects.ForeignId equals foreignObjects.Id
orderby foreignObjects.PropertyYouWantToSortOn
select primaryObjects;
So you have table A which has id of table B as a foreign key and you want to sort table A by the DisplayName column of table B rather than the id of table B?
The only way to achive this is by a join.
SELECT tableA.* FROM tableA INNLER JOIN tableB ORDER BY tableB.DisplayName
In SubSonic2 you can do that, and still be able to update your records if you use the DB.Select(...).ExecuteCollection() method.
I think this should be possible with subsonic3, too.
Howevery, if you don't use the foreign key and the display name is unique, you should just use this value as your foreign key.