We've got a pattern that has worked in the past for insert
statements, but I'm attempting to reuse it for a select
statement and seeing what I'd consider to be unexpected behavior. Assume everything that isn't delineated here is correctly defined; I'm trying to keep this brief, not cover the nitty-gritty details.
So I have an entity (all code is Groovy):
class SomeEntity {
static class COLUMNS {
public static final String COL_1 = 'col_1'
public static final String COL_2 = 'col_2'
// ...elided
}
static class FIELDS {
public static final Field<String> COL_1 = DSL.field(COLUMNS.COL_1, String)
public static final Field<String> COL_2 = DSL.field(COLUMNS.COL_2, String)
// ...elided
}
static List<Field<?>> ENTITY_FIELDS = [FIELDS.COL_1, FIELDS.COL_2]
@Column(name = COLUMNS.COL_1)
String col1
@Column(name = COLUMNS.COL_2)
String col2
}
(This is severely truncated as a minimal example. There are also Boolean
, LocalDateTime
, and BigDecimal
Fields.)
And then a repo like this:
@Repository
class SomeEntityRepo {
DSLContext jooq
SomeEntityRepo(DSLContext jooq) {
this.jooq = jooq
}
SomeEntity findEntityDate(String id) {
return jooq.select(ENTITY_FIELDS)
.from(table)
.where(COL_1.eq(id))
.fetchOneInto(SomeEntity)
}
}
(also truncated, we use constructor injection... there's a join that will happen, etc.)
What I'm seeing is that jOOQ is executing a select * from ...
rather than a select col_1, col_2 from ...
. The minute I move the declaration of ENTITY_FIELDS
into the repository class, it works as expected.
The main issue here is that this should actually end up as select table.col_1, table.col_2 from ...
because there's a join and, of course, both tables have an id
column. I can, of course, define the List<Field>
in the repo class, but I have more than one such list, so it's kind of nice to have them declared (and hidden away) in the entity classes. Also, there's another query that's more join-heavy and ultimate what got me hunting down this issue.
Again, using these entity-defined Lists works just fine for insert
statements, so I'm confused why they are not recognized correctly by the select
. I've read the documentation and noticed that *
is substituted when columns are not recognized, but I've even gone so far as to reduce the entity-defined List to just the ID
field and it still does a select *
.
Am I missing something? Is there a reason for this? Is the behavior caused by jOOQ or is this a Spring issue? This project is using Spring Boot 3.2.2 and jOOQ 3.17.6.
P.S. Update:
I've continued to mess around here and think I may have found a reason for this, but it also boggles my mind. The SomeEntity
class also has this going on:
class SomeEntity {
public static final String TABLE_NAME = 'table_name'
public static final org.jooq.Table ENTITY_TABLE = DSl.table(TABLE_NAME)
public static final org.jooq.Table OTHER_TABLE = DSL.table('other_table')
// and two more for other entities
static class FIELDS {
// previous entries elided
public static final Field<String> OTHER_ID = DSL.field("$ENTITY_TABLE.other_id") // this is not exactly right, but gives the general idea of what we do in many places
// repeat for a handful of other fields.
}
}
If I move those additional entries to the repo class, everything suddenly works as expected (with the ENTITY_FIELDS
list in-use). I can leave the "additional Tables" and it's fine. If I also move the additional tables, it starts doing select *
again.
I'm going to continue to experiment; these additional tables/fields technically don't belong in SomeEntity
, so I'll see if refactoring them, together, works, but I'm not hold my breath given the above information.
This appears to have been due to some random relation/interaction that jOOQ had with a Map
field in the repo class. Originally, I added it as a bit of a shortcut because it somehow allowed jOOQ to correctly assign properties in the .fetchOneInto(ResultObject)
bit.
This was, of course, in an effort to remove the need to implement a fully-defined RecordMapper.map
method. Being that this "result object" has ~111 properties, I really didn't want to have to define every single one...
So, the "shortcut" map was as simple as this:
Map<String, Object> someRandomName = [
(FIELDS.COL_1): 'col_1',
(FIELDS.COL_A): 'col_a'
]
Where FIELDS.COL_A
is defined on a subclass of SomeEntity
. Nothing else was necessary for some "magic" to happen, but as a result of that magic, the ENTITY_FIELDS
List suddenly contained only null
elements.
What's interesting is that the mere existence of this map seems to cause the issue, even when it's not used, even when the proper RecordMapper.map
method is implemented. Even more intriguing is that I can comment COL_1
's mapping out and its list is then non-null, but the COL_A
class's list remains null-populated, and vice versa.