squeryl / squeryl

A Scala DSL for talking with databases with minimum verbosity and maximum type safety

How to build

1. Download or clone repository:
   git clone [email protected]:squeryl/squeryl.git
2. Open a shell in the project's root directory and launch SBT with ./sbt this will fetch the required version of Scala both for SBT itself and for Squeryl.
3. The 'test' command will run the test suite against the minimalist but very complete H2 database. Type 'package' to create a jar in ./target/[scala version].

For more information

Documentation can be found at The Squeryl Website and Scaladoc. Questions can be directed to the Google Group

squeryl's Issues

"Foreing" should be "Foreign"

Schema.scala:

def applyDefaultForeingKeyPolicy(foreingKeyDeclaration: ForeingKeyDeclaration) =
foreingKeyDeclaration.constrainReference

should be

def applyDefaultForeignKeyPolicy(foreignKeyDeclaration: ForeignKeyDeclaration) =
foreignKeyDeclaration.constrainReference

And the problem propagates to the examples that override it. Yes, english is a bastard language, sorry!

No such method error refreshing StatefulManyToMany

Hi Max,

Evaluating a stateful ManyToMany results in the following:

Exception in thread "main" java.lang.NoSuchMethodError: scala.collection.generic.Growable.$plus$plus$eq(Lscala/collection/Traversable;)Lscala/collection/generic/Growable;
at org.squeryl.dsl.StatefulManyToMany.refresh(ManyToMany.scala:189)
at org.squeryl.dsl.StatefulManyToMany.(ManyToMany.scala:185)
at org.squeryl.dsl.ManyToManyRelation$class.rightStateful(ManyToMany.scala:105)
at org.squeryl.dsl.QueryDsl$ManyToManyRelationImpl.rightStateful(QueryDsl.scala:266)

At first I thought the problem was that the HashMap was not mutable, but the exception indicates that it is a Growable. It seems like this method should exist, so I am puzzled.

-Aled

Couldn't make a clone of source

I am new to github and may be doing something wrong and bear with me if so. I am getting the error
" fatal : unable to connect a socket (Invalid argument) " error while trying to clone using the command
git clone git://github.com/max-l/Squeryl.git

"table.column in Nil" causes MySQLSyntaxErrorException with MySQL

The "in" operator assumes that it will not receive an empty list as its argument, and fails when it does.

BigDecimal support

A bug in Session object

object Session {

private val _currentSessionThreadLocal = new ThreadLocal[Option[Session]] {
override def initialValue = None
}

def cleanupResources =
// TODO: Bug here
if(_currentSessionThreadLocal != None)
_currentSessionThreadLocal.get.get.cleanup

}

Support for Hints

Passing column arguments to count

The count function needs to take (optional) column arguments.

For examples, see discussion on Squeryl mailing list: http://groups.google.com/group/squeryl/browse_thread/thread/5be2d1dd8d7c6297/3c47cef42fd7248d#3c47cef42fd7248d

Unsupported fields should be rejected early

Currently Squeryl simply ignores fields of unsupported types in table classes. It should probably throw an exception of some sort for unannotated fields of unsupported types , and provide some sort of annotation to allow fields to exist in table classes that are ignored by Squeryl (like @transient to match the EJB3 persistence annotations).

Support KeyedEntity with auto increment = false

Make InnoDB the default MySql engine for table creation

Equality expression not reversible in OneToMany

http://groups.google.com/group/squeryl/browse_thread/thread/86a444531067a420

Class cast exception on OneToMany query

http://groups.google.com/group/squeryl/browse_thread/thread/6be101be1fd4df67?hl=en_US

CustomTypes constants within expressions don't work

http://groups.google.com/group/squeryl/browse_thread/thread/8b81c1b1750d50ee

Timestamp as query input param gets truncated

Note : no truncation occurs in the mapping to and from the DB

http://groups.google.com/group/squeryl/browse_thread/thread/b1cb87c9bc5b04c1

broken sql generated for tables with non-integer id

When table have non-integer id, generated DML statement include auto_increment.
Example:

class P(override val id:StringField,
                 val descr:StringField) extends KeyedEntity[StringField]

generate

create table P (
   descr varchar(123) not null
    id varchar(123) not null primary key auto_increment
  );

ConstantExpressionNodeList should use Traversable[A] instead of List[A]

Depend only on the lowest common denominator

Support PostgreSQL enum, please :)

Currently, if table.column is of type enum, and generated SQL contains "where table.column = ?" with bound string parameter, PostgreSQL throws:

org.postgresql.util.PSQLException: ERROR: operator does not exist: column = character varying
Hint: No operator matches the given name and argument type(s). You might need to add explicit type casts.

Shrink oracle column aliases and all generated object names to support Oracle's 30 char limit

View.lookup querying 3 times

http://groups.google.com/group/squeryl/browse_thread/thread/8bbebcdbcf97ec27

Support mixed outer and inner joins

If there is at least one outer join, all tables in the from clause with the exception of the first one become "left inner" joined.

schema.printDml must be schema.printDdl

schema.printDml must be schema.printDdl, becouse it pring database definition statements, not
database manipulation statements.

Setting length of String column via annotation doesn't work

Mutable Statements

Statements are immutable by default, this means that even parameters are set once and for all, in some cases it might be desirable to reuse a given statement with new parameters, allowing the preparedStatement to be reused for increased performance.

Add support for DB2

Some of the things the adapter will probably need:

• Boolean type will need to be either char(1) or smallint. Would be nice to be a setting given people have a preference for either one.
• timestamp type is timestamp
• paginated queries will probably need to be overridden. DB2 uses a similar style (if not identical) to Oracle.
• DB2 supports identity columns (similar to SQL Server) but does not use auto_increment. So I'd say to use sequences as per Oracle.
• Will need to get the error numbers for the various failures such as table doesn't exist.
• There will be issues with schemas. Issue 2 would resolve this, but there are workarounds (eg set a default schema in db2cli.ini).

Allow the declaration of schema object prefixes at schema and table level.

After this discussion: http://groups.google.com/group/squeryl/browse_thread/thread/159c423811acdf9c

Besides virtual method you suggested, there's another idea: maybe it would be ok to introduce special annotation to be used like this:

@SchemaName('dict')
object Dict extends Schema {
    val countries = table[Country];
    val currencies = table[Currency];
}
@SchemaName('billing')
object Billing extends Schema {
    val history = table[History];
}

Here I've shown the way I'd use it: each database schema would correspond to its own Schema instance. Expected Squeryl behaviour is this: if @SchemaName annotation is there, then Query referencing Billing.history would generate SQL with "from billing.history alias" instead of "from history alias".

This also requires that multiple Schema instances can be used together in the single Query:

from(Billing.history, Dict.currencies)((h,c) =>
    where(c.id === h.currencyId)
    select((h, c))
)

Schema.printDML broken

Can't find @Column ?

There is a demo in the site:

http://squeryl.org/schema-definition.html

             class Book(val id: Long,
   13           var title: String,
   14           @Column("AUTHOR_ID") // the default 'exact match' policy can be overriden
   15           var authorId: Long,
   16           var coAuthorId: Option[Long]) {

Note the line 14, there is a Column annotation. But I can't find such a annotation in the release code. Where is wrong?

Add dynamic *and* pagination queries to the test suite

Exception thrown in outer join with subquery member

http://groups.google.com/group/squeryl/browse_thread/thread/ba310952e3097e95

Feature Request: Allow custom database types for certain columns

Related discussion: http://groups.google.com/group/squeryl/browse_thread/thread/c5866d1a44853700

It would be helpful to allow custom database types for certain columns, i.e. "timestamp with time zone" instead of "timestamp"
Maxime's current idea is:
"I plan to delegate this
method (databaseTypeFor) to the Schema trait, there will two levels where one
can override the default DB type

1. Schema wide, by overriding at the Schema trait
2. Per field, i.e. with an annotation

Of course (2) would have precedence over (1)"

At the moment, a workaround would be (for the timestamp example):

new PostgresSqlAdapter(...)  {
  override def databaseTypeFor(fmd: FieldMetaData) =   
    if(fmd.wrappedFieldType.isInstanceOf[Timestamp])
      "timestamp with time zone"
    else
      super.databaseTypeFor(fmd)
}

Support index and uniqueness constraints for Schema creation

http://groups.google.com/group/squeryl/browse_thread/thread/c118fcb9c18d7f9b

Feature request: POSIX regular expression search operator

All databases currently supported by Squeryl have regular expression search in some way or other. It would be useful for Squeryl to have a "regex" operator that could be used, for example, in the following manner:

[...](n => where%28%28n.name%29 regex%28"^N"%29%29 select%28n%29) [...]

See also http://groups.google.com/group/squeryl/browse_thread/thread/1b41c9f4e46c10c/8d100b5658d9f3c0#8d100b5658d9f3c0 (where you can find out how to implicitly add such a feature yourself, without touching the Squeryl code)

Kind regards,
/nikolaj

Update with composite primary key fails

http://groups.google.com/group/squeryl/browse_thread/thread/41ee169f0f3ead92

Make transparent retries and timeouts

Per http://groups.google.com/group/squeryl/browse_thread/thread/15f0d52ea2288ee2
I'm envious of the decorators that querulous has for common system concerns for database connection and query resilience. It would be great if Squeryl also provided some help like that.

FieldReferenceLinker._lastAccessedFieldReference bleed-through

http://groups.google.com/group/squeryl/browse_thread/thread/80a0c14e72d5bca/9f1c3d4113161245?lnk=gst&q=bleed#9f1c3d4113161245

Reported by Alec Wysoker :

I mentioned previously that sometimes Squeryl produced SQL queries that were
not valid, but I couldn't come up with a good reproduction procedure. I
still don't have a great reproduction procedure independent of my code base,
but after much debugging I have gained some understanding of what is going
on. There is an attribute FieldReferenceLinker._lastAccessedFieldReference.
It looks like this is supposed to be used, then cleared when it is done
being used. It doesn't appear to be getting cleared in all cases. What I
see is that sometimes a value stored in the attribute from a previous
operation bleeds through into the next query, and messes up the query.

I have seen this when using relations like this:

lazy val notificationStatuses =
MySchema.workflowToNotificationStatus.left(this)

I have worked around this problem by doing this:

lazy val notificationStatuses = {
MySchema.clearFieldReferenceLinker
MySchema.workflowToNotificationStatus.left(this)
}

Where MySchema.clearFieldReferenceLinker is

def clearFieldReferenceLinker = (new YieldInspection).cleanUp

I haven't wrapped my head around how this is supposed to work, or if what
I'm doing is the right thing.

Default string length 123?

In Schema.scala

/**

• @return the default database storage (column) length for String columns for this Schema,
• Can be overridden by the Column Annotation ex.: Column(length=256)
• default is 128
  */
  def defaultLengthOfString = 123

The default varchar length should be 128 right?

private member with no public accessor are considered tansient by squeryl

http://groups.google.com/group/squeryl/browse_thread/thread/f6b3ea7e1e7ad8e0

The repository uses tainted line endings

Every Git repository needs to use \n, also known as line feed characters, for its line endings in every text file in the repository, or else there will be problems with merges and/or plain-text patches. The issue is outlined in more detail in the GitHub help pages.

As you all know, Windows uses \r\n instead for its line endings (carriage return + line feed), and it seems like Maxime hasn't thought about this issue when committing, so the majority (but not all!) of the files in the repository now use the Windows-style line endings. I stumbled upon this fact when:

1. Emacs complained about it ("DOS emulation mode for Git repository").
2. Kate complained about it (Incessant "Do you want to convert this DOS text file?" questions).
3. Git-Flow complained about it (Failed to do a feature branch merge).
4. Git-Gui complained about it (There were ^M warnings all over the place).

The fix for this is simple:

1. Maxime and other Windows-users need to issue the command git config --global core.autocrlf true so that files get \r\nendings when they are checked out, and get \n endings again when they are committed back in.
2. I've made a commit that fixes the issue in dflemstr/Squeryl@e13339c. It looks like every file in the repository has been completely rewritten, but it's only the line endings that are changing. I didn't want to push this to max-l's repository yet, since he hasn't activated auto-CRLF conversion yet.
3. People who have created forks of the repository should pull in this new change as soon as possible, so that they don't make changes that also use CRLF-endings (because most editors will continue to use CRLF-endings if they open a file with CRLF-endings)
4. From time to time, a check should be made that no more CRLF endings have crept in. I could do that until further notice.

SqlServer support

UUID support

Postgres has native UUID support. Other databases can use binary[16] types usually. It would be nice to have support for this natively within Squeryl. I have code that uses UUIDs as that ID for objects and would like to persist them out to a database without too much pain.

PostgreSQL PK assignment after Table[T].insert(T).

I got PostgreSQL error:

Exception while executing statement :
insert into Article (id, url, when_last_modified, is_draft, annotation, text, sortOrder, title, when_created) values (nextval('seq_Article'),?,?,?,?,?,?,?,?)
....
Caused by: org.postgresql.util.PSQLException: ERROR: relation "seq_article" does not exist

A sequence name PostgreSQL 8.4 generated for "id serial" field is "article_id_seq", not "seq_article". I suppose there should be Table annotation or so, to provide correct sequence name.

Timestamp type becomes 'date' in MySql, should be 'datetime'

Problems creating indices: missing single-column indices

http://groups.google.com/group/squeryl/browse_thread/thread/763b5295e176f0

Support string queries (JDBC wrapper)

It would be nice to have a thin wrapper around JDBC for things that are not (yet) possible to do in squeryl, it could be maybe similar to querolus.

transaction(session){ //this should work in a transaction of the session created by squeryl
    val items = for(row <- select("stored","param1",1,2,3) yield {
        Item(row.getInt("id"),row.getInt("count"))
    }
} 

Other use cases would be helpers for stored procedure execution with signatures like these (psql output):

Result data type	Argument data types
refcursor	curs refcursor, name text, filter text
record	id integer, name text, maxcount integer, OUT resultid integer, OUT message text

Spelling errors in identifiers in the code

While the foreign stuff was fixed (thanks!), these things remain as spelling errors in the code:

• fieldMataData
• Yeld
• Correspondance
• Agregate
• UniqueIdInAliaseRequired
• colAnotations
• courseAssigments
• SchoolDb2MetableRelations
• inerpolater
• Interpolater
• Convertion

I don't know what's public and what isn't ... Some of them are just in test code also.

FieldMetadata.isAutoIncremented is not set in timely fashion

I am trying to insert into a database for which the schema was created by a previous program. For the first element that I insert, the isAutoIncremented attribute is not set to true. It appears that this is because PosoMetaData.primaryKey is a lazy val, but it has a side-effect of setting isAutoIncremented on the primary key's FieldMetaData.

I don't know why this is a lazy val, but one way to addres it would be in to access posoMetaData.primaryKey in Table.insert prior to calling _dbAdapter.writeInsert.

Oracle busting the 30 char limit when using schema prefix

sql case statemens

implement it !

Add support for HyperSQL (HSQLDB)

Should be straight forward. The SQL syntax is very similar to DB2