An alternative API for filtering data with Spring MVC & Spring Data JPA.
A thorough introduction and the original rationale behind this component can be found my blog: http://blog.kaczmarzyk.net/2014/03/23/alternative-api-for-filtering-data-with-spring-mvc-and-spring-data/. In this file you can find a summary of all the current features and some API examples.
You can also take a look on working Spring Boot apps that use this library:
- for Spring Boot 3: https://github.com/tratif/specification-arg-resolver-example-springboot3
- this includes demonstration of Native Images for GraalVM
- for Spring Boot 2: https://github.com/tkaczmarzyk/specification-arg-resolver-example.
- Basic usage -- quick start with the lib
- Simple specifications -- basic specs, such as
Equal
,Like
,GreaterThan
etc. - Combining specs
- @And -- combining simple specs with
and
keyword - @Or -- combining simple specs with
or
keyword - Nested conjunctions and disjunctions
- @And -- combining simple specs with
- Join -- filtering by attributes of joined entities
- Join fetch -- initializing lazy associations
- Advanced HTTP parameter handling
- Static parts of queries -- adding static (not bound to HTTP params) predicates such as
deleted = false
- Default value of queries -- providing a fallback value when HTTP parameter is not present
- Annotated specification interfaces -- resolving specifications from annotated interfaces
- Handling different field types -- handling situations when provided parameter is of different type than the field (e.g.
"abc"
sent against an integer field) - Path Variable support -- using uri fragments (resolvable with Spring's
@PathVariable
annotation) in specifications - Json Request Body support -- using json in request body to get parameters for specification
- Type conversions for HTTP parameters -- information about supported type conversions (i.e. ability to convert HTTP parameters into Java types such as
LocalDateTime
, etc.) and the support of defining custom converters - Locale support -- information about
Locale
configuration for case-insensitive matching - SpEL support -- information about Spring Expression Language support
- Swagger support -- information about support for generation of swagger documentation
- Building specifications outside the web layer
- Compatibility notes -- information about older versions compatible with previous Spring Boot and Java versions
- Download binary releases -- Maven artifact locations
The following HTTP request:
GET http://myhost/api/customers?firstName=Homer
can be handled with the following controller method:
@RequestMapping(value = "/customers", params = "firstName")
public Iterable<Customer> findByFirstName(
@Spec(path = "firstName", spec = Like.class) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
which will result in the following JPA query:
select c from Customer c where c.firstName like '%Homer%'
Alternatively you can annotate an interface:
@Spec(path="firstName", params="name", spec=Like.class)
public interface NameSpec extends Specification<Customer> {
}
and then use it as a controller parameter without any further annotations.
All you need to do is to wire SpecificationArgumentResolver
into your application. Then you can use @Spec
and other annotations in your controllers. SpecificationArgumentResolver
implements Spring's HandlerMethodArgumentResolver
and can be plugged in as follows:
@Configuration
@EnableJpaRepositories
public class MyConfig implements WebMvcConfigurer {
@Override
public void addArgumentResolvers(List<HandlerMethodArgumentResolver> argumentResolvers) {
argumentResolvers.add(new SpecificationArgumentResolver());
}
...
}
The library converts HTTP parameters (strings) into most popular Java types such as dates, numbers, and enums. In case of a need for some additional conversion, please see how to configure specification argument resolver with If you need to use additional converters please see custom converters section.
Use @Spec
annotation to automatically resolve a Specification
argument of your controller method. @Spec
has path
property that should be used to specify property path of the attribute of an entity, e.g. address.city
. By default it's also the name of the expected HTTP parameter, e.g. GET http://myhost?address.city=Springfield
.
Use spec
attribute of the annotation to specify one of the filtering strategies listed in the subsections below (e.g. Like, Equal, In)
For multi value filters like: In.class
, NotIn.class
there are two ways of passing multiple arguments. The first way is passing the same HTTP parameter multiple times as follows:
GET http://myhost/customers?gender=MALE&gender=FEMALE
The second way is the use paramSeparator
attribute of @Spec
, which determines the argument separator (can be specified only for query parameters).
For example the following controller method:
@RequestMapping(value = "/customers", params = "genderIn")
@ResponseBody
public Object findCustomersByGender(
@Spec(path = "gender", params = "genderIn", paramSeparator = ",", spec = In.class) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
will handle GET http://myhost/customers?gender=MALE,FEMALE
in exactly the same way as GET http://myhost/customers?gender=MALE&gender=FEMALE
(as one parameter with two values ["MALE","FEMALE"]
). Without specifying paramSeparator
param gender=MALE,FEMALE
will be processed as single value: ["MALE,FEMALE"]
.
Filters using JPAQL like
expression. It adds a wildcard %
at the beginning and the end of the actual value, e.g. (..) where firstName like %Homer%
.
Usage: @Spec(path="firstName", spec=Like.class)
.
There are also other variants which apply the wildcard only on the beginning or the ending of the provided value: StartingWith
and EndingWith
.
The negated version is available: NotLike
which executes queries such as (..) where firstName not like %Homer%
Works as Like
, but the query is also case-insensitive.
Usage: @Spec(path="firstName", spec=LikeIgnoreCase.class)
.
There are also other variants which apply the wildcard only on the beginning or the ending of the provided value: StartingWithIgnoreCase
and EndingWithIgnoreCase
.
Locale settings is important for case-insensitive searches. Please check the Locale support section for details.
A negation for this specification is also available: NotLikeIgnoreCase
.
Compares an attribute of an entity with the value of a HTTP parameter (exact match). E.g. (..) where gender = FEMALE
.
Supports multiple data types: numbers, booleans, strings, dates, enums.
Usage: @Spec(path="gender", spec=Equal.class)
.
The default date format used for temporal fields is yyyy-MM-dd
. It can be overriden with a configuration parameter (see LessThan
below). If for date-time formats which store also time (LocalDateTime
, OffsetDateTime
, Instant
and Timestamp
) only the date is provided, then the default time value (midnight - UTC time for OffsetDateTime
) will be used for checking equality. To include all results within particular date (day) use EqualDay
specification.
A negation for this specification is also available: NotEqual
.
Works as Equal
, but the query is also case-insensitive, could be used for fields of type: String
, Enum
.
Locale settings is important for case-insensitive searches. Please check the Locale support section for details.
A negation for this specification is also available: NotEqualIgnoreCase
.
Compares an attribute of an entity with multiple values of a HTTP parameter. E.g. (..) where gender in (MALE, FEMALE)
.
Supports multiple data types: numbers, booleans, strings, dates, enums.
Usage: @Spec(path="gender", spec=In.class)
.
HTTP request example:
GET http://myhost/customers?gender=MALE&gender=FEMALE
or if paramSeparator
is specified (eg. @Spec(path="gender", paramSeparator=",", spec=In.class)
):
GET http://myhost/customers?gender=MALE,FEMALE
The default date format used for temporal fields is yyyy-MM-dd
. It can be overridden with a configuration parameter (see LessThan
below).
A negation for this specification is also available: NotIn
.
Filters using is null
or is not null
, depending on the value of the parameter passed in. A value of true
will filter for is null
, and a value of false
will filter for is not null
.
The data type of the field specified in path
can be anything, but the HTTP parameter must be a Boolean. You should use params
attribute to make it clear that the parameter is filtering for null values.
Usage: @Spec(path="activationDate", params="activationDateNull" spec=Null.class)
.
If you want the query to be static, i.e. not depend on any HTTP param, you can use IsNull
or IsNotNull
specifications. Alternatively, you can use Null
with constVal
attribute of Spec
annotation:
For example @Spec(path="nickname", spec=Null.class, constVal="true")
will always add nickname is null
to the query.
An inversion of Null
described above, for better readability in some scenarios.
For example, consider a deletedDate
field which is null when the entity is not deleted, and vice-versa. Then, you can introduce this mapping:
@Spec(path="deletedDate", params="isDeleted", spec=NotNull.class)
to handle HTTP requests such as:
GET http://myhost/customers?isDeleted=true
GET http://myhost/customers?isDeleted=false
to return deleted (deletedDate
not null) and not deleted (deltedDate
null) respectively.
Filters with is null
where clause for particular field defined under path
in Spec
annotation. Does not require any http-parameters to be present, i.e. represents constant part of the query. The same effect can be achieved with Null
specification and @Spec.constVal
set to true
. IsNull
is just a convenience class that can make the code more explicit.
For example, consider nickName
field. Then, you can introduce the following mapping:
@Spec(path="nickName", spec=IsNull.class)
to handle HTTP requests such as:
GET http://myhost/customersWithoutNickname
to return customers without nickname (nickName
field with null
value).
A negation for this specification is also available: IsNotNull
.
Filters using is empty
or is not empty
, depending on the value of the parameter passed in. A value of true
will filter for is empty
and a value of false
will filter for is not empty
collections (e.g. where customer.orders is empty
).
Supports collection data types specified under path
in Spec
annotation. Provided HTTP parameter must be a Boolean. You should use params attribute to make it clear that the parameter is filtering for results with empty (not empty) particular collection.
Usage: @Spec(path="orders", params="emptyOrders", spec=Empty.class)
.
If you want the query to be static, i.e. not depend on any HTTP param, you can use IsEmpty
or IsNotEmpty
specifications. Alternatively, you can use Empty
with constVal
attribute of Spec
annotation.
For example @Spec(path="orders", spec=Empty.class, constVal="true")
will filter for results with empty orders
collection.
A negation for this specification is also available: NotEmpty
.
Filters with is empty
where-clause for collections that are defined under path
in Spec
annotation (e.g. where customer.orders is empty
). Does not require any http-parameters to be present, i.e. represents constant part of the query.
For example, for orders
table which contains association to customers
table (one customer can have multiple orders) the following mapping can be introduced:
@Spec(path="orders", spec=IsEmpty.class)
to handle HTTP request such as:
GET http://myhost/customersWithoutOrders
to return customers without orders (with empty collection of orders).
A negation for this specification is also available: IsNotEmpty
.
Filters using true
or false
for a boolean type field, depending on the value of the parameter passed in (e.g. where customer.gold = true
). This can be also achieved with Equal
. True
is a convenience-class to emphasize boolean attributes.
The HTTP parameter and data type of the corresponding field specified in path
must be a Boolean.
For example, consider gold
field which indicates special customer
. Then, you can introduce the following mapping:
@Spec(path="gold", params = "golden", spec=True.class)
to handle HTTP requests such as:
GET http://myhost/customers?golden=true
to return golden customers (with gold
field set to true
).
If you want the query to be static, i.e. not depend on any HTTP param, you can use IsTrue
or IsFalse
specifications. Alternatively, you can use True
with constVal
attribute of Spec
annotation.
For example @Spec(path="gold", spec=True.class, constVal="true")
will filter for golden customers.
A negation for this specification is also available: False
.
Filters with true
value of particular field defined under path
in Spec
annotation. Does not require any http-parameters to be present, i.e. represents constant part of the query. The same effect can be achieved with Equal
specification and @Spec.constVal
set to true
. IsTrue
is just a convenience class that can make the code more explicit.
For example, consider gold
field which indicates special customer
. Then, you can introduce the following mapping:
@Spec(path="gold", spec=IsTrue.class)
to handle HTTP requests such as:
GET http://myhost/goldenCustomers
to return all special customers (with gold
field set to true
).
A negation for this specification is also available: IsFalse
.
Checks if the value passed as HTTP parameter is a member of a collection attribute of an entity (defined under path
in @Spec
annotation).
Supports collections of multiple data types: numbers, booleans, strings, dates, enums. Only @ElementCollection
annotation with basic data types is supported (collections of Embeddable
objects or @OneToMany
relationships with other entities are not supported).
For example, let's assume a Customer
entity with luckyNumbers
collection attribute (one customer can have multiple lucky numbers). The following mapping can be introduced:
@Spec(path="luckyNumbers", params = "luckyNumber", spec=IsMember.class)
to handle HTTP requests such as:
GET http://myhost/customers?luckyNumber=777
to return customers with particular lucky number.
A negation for this specification is also available: IsNotMember
.
Filters using a comparison operator (>
, >=
, <
or <=
). Supports multiple field types: strings, numbers, booleans, enums, dates. Field types must be Comparable (e.g, implement the Comparable interface); this is a JPA constraint.
Usage: @Spec(path="creationDate", spec=LessThan.class)
.
For temporal values, the default date format is yyyy-MM-dd
. You can override it by providing a config value to the annotation: @Spec(path="creationDate", spec=LessThan.class, config="dd-MM-yyyy")
. When the date format is specified without the time part for datetime types, the missing time values will be filled with zeros.
NOTE: comparisons are dependent on the underlying database.
- Comparisons of floats and doubles (especially floats) may be incorrect due to precision loss.
- Comparisons of booleans may be dependent on the underlying database representation.
- Comparisons of enums will be of their ordinal or string representations, depending on what you specified to JPA, e.g.,
@Enumerated(EnumType.STRING)
,@Enumerated(EnumType.ORDINAL)
or the default (@Enumerated(EnumType.ORDINAL)
)
Filters by checking if a comparable field of an entity is in the provided range. E.g. (..) where creation date between :after and :before
.
It requires 2 HTTP parameters (for lower and upper bound). You should use params
attribute of the @Spec
annotation, i.e.: @Spec(path="registrationDate", params={"registeredAfter","registeredBefore"}, spec=Between.class)
. The corresponding HTTP query would be: GET http://myhost/customers?registeredAfter=2014-01-01®isteredBefore=2014-12-31
.
NOTE: comparisons are dependent on the actual type and the underlying database (see the explanation for GreaterThan
above).
You can configure the date/datetime pattern as with LessThan
described above.
Filters using comparison operators (>
, <
). Supports date-type fields. Compares current db timestamp to the date-type value of field passed in path
.
E.g. InTheFuture => where customer0_.date_of_next_special_offer > localtimestamp()
, InThePast => where customer0_.date_of_next_special_offer < localtimestamp()
Usage: @Spec(path="dateOfTheNextOffer", spec=InTheFuture.class)
.
Supports date-type fields. Matches the day part of a date-time attribute, ignoring the time. Filters with range within one day from 00:00:00
(inclusive) to 00:00:00
of the next day (exclusive). When datetime format with a time part will be specified and provided, then the time part will be ignored. For example, providing 2022-12-20T08:45:57
parameter for LocalDateTime
field will result with filtering with range from 2022-12-20T00:00:00
(inclusive) to 2022-12-21T00:00:00
(exclusive).
Usage: @Spec(path="lastOrderTime", spec=EqualDay.class)
.
For types that store just date (without time) like LocalDate
the specification behaves the same as Equal
.
For more information related to date-types see Supported Conversions and Date Formats.
You can combine the specs described above with or
& and
. Remember that by default all of the HTTP params are optional. If you want to make all parts of your query required, you must state that explicitly in @RequestMapping
annotation (see above).
Usage:
@RequestMapping("/customers")
public Object findByName(
@And({
@Spec(path="registrationDate", params="registeredBefore", spec=DateBefore.class),
@Spec(path="lastName", spec=Like.class)}) Specification<Customer> customerSpec) {
return customerRepo.findAll(customerSpec);
}
would handle requests like GET http://myhost/customers?registeredBefore=2015-01-18&lastName=Simpson
and execute queries like: select c from Customer c where c.registrationDate < :registeredBefore and c.lastName like '%Simpson%'
.
Usage:
@RequestMapping("/customers")
public Object findByName(
@Or(
@Spec(path="firstName", params="name", spec=Like.class),
@Spec(path="lastName", params="name", spec=Like.class)) Specification<Customer> customerNameSpec) {
return customerRepo.findAll(customerNameSpec);
}
would handle requests like GET http://myhost/customers?name=Mo
and execute queries like: select c from Customer c where c.firstName like '%Mo%' or c.lastName like '%Mo'
.
You can put multiple @And
inside @Disjunction
or multiple @Or
inside @Conjunction
. @Disjunction
joins nested @And
queries with 'or' operator. @Conjunction
joins nested @Or
queries with 'and' operator. For example:
@RequestMapping("/customers")
public Object findByFullNameAndAddress(
@Conjunction({
@Or(@Spec(path="firstName", params="name", spec=Like.class),
@Spec(path="lastName", params="name", spec=Like.class)),
@Or(@Spec(path="address.street", params="address", spec=Like.class),
@Spec(path="address.city", params="address", spec=Like.class))
}) Specification<Customer> customerSpec) {
return customerRepo.findAll(customerSpec);
}
would handle requests like GET http://myhost/customers?name=Sim&address=Ever
and execute queries like select c from Customer c where (c.firstName like '%Sim%' or c.lastName like '%Sim%') and (c.address.street like '%Ever%' or c.address.city like '%Ever%')
.
You must use @Conjunction
and @Disjunction
as top level annotations (instead of regular @And
and @Or
) because of limitations of Java annotation syntax (it does not allow cycle in annotation references).
You can join nested @And
and @Or
queries with simple @Spec
, for example:
@RequestMapping("/customers")
public Object findByFullNameAndAddressAndNickName(
@Conjunction(value = {
@Or(@Spec(path="firstName", params="name", spec=Like.class),
@Spec(path="lastName", params="name", spec=Like.class)),
@Or(@Spec(path="address.street", params="address", spec=Like.class),
@Spec(path="address.city", params="address", spec=Like.class))
}, and = @Spec(path="nickName", spec=Like.class) Specification<Customer> customerSpec) {
return customerRepo.findAll(customerSpec);
}
@RequestMapping("/customers")
public Object findByLastNameOrGoldenByFirstName(
@Disjunction(value = {
@And({@Spec(path="golden", spec=Equal.class, constVal="true"),
@Spec(path="firstName", params="name", spec=Like.class)})
}, or = @Spec(path="lastName", params="name", spec=Like.class) Specification<Customer> customerSpec) {
return customerRepo.findAll(customerSpec);
}
You can use @Join
annotation to perform joins and then filter by attributes of joined entities. For example, let's assume the following entities:
@Entity
public class Customer {
// other fields omitted for brevity
@OneToMany(mappedBy = customer)
private Collection<Order> orders;
}
@Entity
public class Order {
// other fields omitted for brevity
@ManyToOne
private Customer customer;
private String itemName;
}
If you want to find all customers who ordered pizza, you can do the following:
@RequestMapping("/customers")
public Object findByOrderedItem(
@Join(path= "orders", alias = "o") // alias specified for joined path
@Spec(path="o.itemName", params="orderedItem", spec=Like.class) // alias used in regular spec definition
Specification<Customer> customersByOrderedItemSpec) {
return customerRepo.findAll(customersByOrderedItemSpec);
}
The default join type is LEFT
. You can use type
attribute of the annotation to specify different value.
Using @Join
annotation makes the query distinct by default. While it is the best approach for most of the cases, you can override it by using distinct
attribute of the annotation. Warning for Hibernate users: from Hibernate version 6 onwards, all queries are distinct anyway and there is no way to change it.
You can specify multiple different joins, for example:
@RequestMapping("/customers")
public Object findByOrderedOrFavouriteItem(
@Join(path = "orders", alias = "o")
@Join(path = "favourites", alias = "f")
@Or({
@Spec(path="o.itemName", params="item", spec=Like.class),
@Spec(path="f.itemName", params="item", spec=Like.class)}) Specification<Customer> customersByItem) {
return customerRepo.findAll(customersByItem);
}
Multi-level joins are supported. To create multi-level join you should specify multiple joins in which join path contains alias of another join with higher priority.
For example, let's assume the following entities:
@Entity
class Customer {
// other fields omitted for brevity
@OneToMany(mappedBy = "customer")
private Set<Order> orders;
}
@Entity
class Order {
// other fields omitted for brevity
@ManyToOne
private Customer customer;
@ManyToMany(fetch = FetchType.LAZY)
private Set<ItemTag> tags;
...
}
@Entity
public class ItemTag {
// other fields omitted for brevity
private String name;
...
}
If you want to find all customers who ordered item tagged as #snacks
, you can do the following:
@RequestMapping(value = "/findCustomersByOrderedItemTag")
@PostMapping
public Object findCustomersByOrderedItemTag(
@Join(path = "orders", alias = "o")
@Join(path = "o.tags", alias = "t")
@Spec(path = "t.name", params = "tag", spec = Equal.class) Specification<Customer> spec) {
return customerRepo.findAll(spec, Sort.by("id"));
}
Annotations are processed sequentially, the order must be kept!
Following spec is invalid:
@RequestMapping(value = "/findCustomersByOrderedItemTag")
@PostMapping
public Object findCustomersByOrderedItemTag(
@Join(path = "o.tags", alias = "t") // "o" alias will be not exist during processing this @Join
@Join(path = "orders", alias = "o")
@Spec(path = "t.name", params = "tag", spec = Equal.class) Specification<Customer> spec) {
return customerRepo.findAll(spec, Sort.by("id"));
}
You can use join annotations with custom annotated specification interfaces.
Inner joins are always evaluated eagerly, even if there is no filtering applied on the joined part (e.g. due to a missing HTTP parameter associated with the specification). This is due to the fact, that inner joins may narrow the result set (entities with empty associations will be ommited). Warning: this behaviour was different in pre 3.X version, see CHANGELOG.md for more details if you are still using 2.X.
Left and right joins are lazily evaluated for distinct queries, i.e. if the HTTP parameter associated with the specification is not present, join will not be performed (as it would be meaningless). For non-distinct queries evaluation will be eager, as such join may expand the result set (duplicated entities might be returned in the result set). Again, this does not matter for Hibernate 6+ users.
You can use @JoinFetch
annotation to specify paths to perform fetch join on. For example:
@RequestMapping("/customers")
public Object findByCityFetchOrdersAndAddresses(
@JoinFetch(paths = { "orders", "addresses" })
@Spec(path="address.city", params="town", spec=Like.class) Specification<Customer> customersByCitySpec) {
return customerRepo.findAll(customersByCitySpec);
}
As with @Join
, the use of @JoinFetch
makes the query distinct by default.
The default join type is LEFT
. You can use joinType
attribute of the annotation to specify different value. You can specify multiple different joins, for example:
@RequestMapping("/customers")
public Object findByCityFetchOrdersAndAddresses(
@JoinFetch(paths = "orders")
@JoinFetch(paths = "addresses", joinType = JoinType.INNER)
@Spec(path="address.city", params="town", spec=Like.class) Specification<Customer> customersByCitySpec) {
return customerRepo.findAll(customersByCitySpec);
}
You can use join annotations with custom annotated specification interfaces.
Multi-level fetch join is supported. To create multi-level fetch join you should specify multiple fetch joins in which join path contains alias of another fetch join with higher priority.
For example, let's assume the following entities:
@Entity
class Customer {
// other fields omitted for brevity
@OneToMany(mappedBy = "customer")
private Set<Order> orders;
}
@Entity
class Order {
// other fields omitted for brevity
@ManyToOne
private Customer customer;
@ManyToMany(fetch = FetchType.LAZY)
private Set<ItemTag> tags;
...
}
@Entity
public class ItemTag {
// other fields omitted for brevity
private String name;
...
}
If you want to find all customers who ordered item with given tag name and fetch additional nested attributes (entities) to avoid SELECT N+1 Problem
you can do the following:
@RequestMapping(value = "/findCustomersWhoOrderedItemWithGivenTag")
@PostMapping
public Object findCustomers(
@JoinFetch(paths = "orders", alias = "o")
@JoinFetch(paths = "o.tags", alias = "t")
@Spec(path = "t.name", params = "tagName", spec = Equal.class) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
The same as in case of multi-level joins, annotations are processed sequentially, the order must be kept!
Please remember that:
- Join fetch path can use only aliases of another fetch joins.
- Join path can use only aliases of another joins.
Following combinations are forbidden:
-
join path which uses join fetch alias
@RequestMapping(value = "/findCustomersByOrderedItemTag", params = { "tagName" }) @PostMapping public Object findCustomers( @JoinFetch(paths = "orders", alias = "o") // Wrong, 'o' defined in join fetch tried to be used in a regular join @Join(path = "o.tags", alias = "t") @Spec(path = "t.name", params = "tagName", spec = Equal.class) Specification<Customer> spec) { return repository.findAll(spec); }
-
join fetch path which uses join alias
@RequestMapping(value = "/findCustomersByOrderedItemTag", params = { "tagName" }) @PostMapping public Object findCustomers( @Join(path = "orders", alias = "o") // Wrong, 'o' defined in a join tried to be used in join fetch @JoinFetch(paths = "o.tags", alias = "t") @Spec(path = "t.name", params = "tagName", spec = Equal.class) Specification<Customer> spec) { return repository.findAll(spec); }
If the same alias is defined both for
@Join
and@JoinFetch
, the join alias will be used during specification resolving.@RequestMapping(value = "/findCustomersByOrderedItemTag", params = { "tagName" }) @PostMapping public Object findCustomers( @JoinFetch(paths = "orders", alias = "o") @JoinFetch(paths = "o.tags", alias = "t") @Join(path = "orders", alias = "o") @Join(path = "o.tags", alias = "t") //'t' refers to third join - @Join(path = "o.tags", alias = "t") @Spec(path = "t.name", params = "tagName", spec = Equal.class) Specification<Customer> spec) { return repository.findAll(spec); }
Using join and join fetch on the same path should be avoided, otherwise the same table will be joined twice (
orders
andtags
in above example). The above example is an anti-pattern and should never be followed in the production code.
If the HTTP parameter is not present, the resolved Specification
will be null
. It means no filtering at all when passed to a repository. If you want to make the parameter non-optional, you should use standard Spring MVC annotations, e.g. @RequestMapping(params={"firstName"})
.
By default, the expected HTTP parameter is the same as the property path. If you want them to differ, you can use params
attribute of @Spec
. For example this method:
@RequestMapping("/customers")
public Object findByCity(
@Spec(path="address.city", params="town", spec=Like.class) Specification<Customer> customersByCitySpec) {
return customerRepo.findAll(customersByCitySpec);
}
will handle GET http://myhost/customers?town=Springfield
as select c from Customer c where city.address like '%Springfield%'
.
If you don't want to bind your Specification to any HTTP parameter, you can use constVal
attribute of @Spec
. For example:
@Spec(path="deleted", spec=Equal.class, constVal="false")
will always produce the following: where deleted = false
. It is often convenient to combine such a static part with dynamic ones using @And
or @Or
described below.
Support for SpEL expression and property placeholders in constVal
could be enabled in following way:
- Configure
SpecificationArgumentResolver
by passing AbstractApplicationContext in constructor - Set attribute
valueInSpEL
value totrue
Configuration example:
@Autowired
AbstractApplicationContext applicationContext;
@Override
public void addArgumentResolvers(List<HandlerMethodArgumentResolver> argumentResolvers) {
argumentResolvers.add(new SpecificationArgumentResolver(applicationContext));
}
Usage example #1:
@RequestMapping(value = "/customers")
@ResponseBody
public Object findCustomersBornInTheFuture(
@Spec(path = "birthDate", constVal = "#{T(java.time.LocalDate).now()}", valueInSpEL = true, spec = GreaterThanOrEqual.class) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
Usage example #2:
@RequestMapping(value = "/customers")
@ResponseBody
public Object findCustomersWithTheLastNameSimpson(
@Spec(path = "lastName", constVal = "${search.default-params.lastName}", valueInSpEL = true, spec = Equal.class) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
application.properties
search.default-params.lastName=Simpson
Defined in constVal
SpEL expression should be able to be evaluated to java.lang.String
.
When the request parameter is not provided or has an empty value, you can use defaultVal
atatribute of @Spec
to provide a value to fallback to. For example this controller method:
@RequestMapping("/users")
public Object findByRole(
@Spec(path="role", spec=Equal.class, defaultVal="USER") Specification<User> spec) {
return userRepo.findAll(spec);
}
Would handle request such as GET /users
with the following query: select u from Users u where u.role = 'USER'
.
Supplying constVal
implicitly sets defaultVal
to empty.
Support for SpEL expression and property placeholders in defaultVal
could be enabled in the same way as for constVal
You can annotate a custom interface that extends Specification
, eg.:
@Or({
@Spec(path="firstName", params="name", spec=Like.class),
@Spec(path="lastName", params="name", spec=Like.class)
})
public interface FullNameSpec extends Specification<Customer> {
}
It can be then used as a controller parameter without further annotations, i.e.:
@RequestMapping("/customers")
@ResponseBody
public Object findByFullName(FullNameSpec spec) {
return repository.findAll(spec);
}
When such parameter is additionally annotated, the both specifications (from the interface and the parameter annotations) are joined with 'and' operator. For example you can define a base interface like this:
@Spec(path="deleted", spec=Equal.class, constVal="false")
public interface NotDeletedEntitySpec<T> extends Specification<T> {}
and then use it as a foundation for you controller as follows:
@RequestMapping("/customers")
@ResponseBody
public Object findNotDeletedCustomerByLastName(
@Spec(path="lastName", spec=Equal.class) NotDeletedEntitySpec<Customer> spec) {
return repository.findAll(spec);
}
to execute queries such as select c from Customer c where c.deleted = false and c.lastName like %Homer%
.
Specifications are resolved from all parent interfaces and combined with and
. As an example, let's consider the following interfaces:
@Spec(path = "deleted", constVal = "false", spec = Equal.class)
public interface NotDeletedSpec extends Specification<Customer> {}
@Spec(path = "firstName", spec = Equal.class)
public interface FirstNameSpec extends NotDeletedSpec {}
FirstNameSpec
extends NotDeletedSpec
, so their specifications will be combined with and
, i.e. a controller method like this:
@RequestMapping("/customers")
@ResponseBody
public Object findNotDeletedCustomersByFirstName(FirstNameSpec spec) {
return repository.findAll(spec);
}
will accept HTTP requests such as GET /customers?firstName=Homer
and execute JPA queries such as where firstName = 'Homer' and deleted = false
.
Consider a field age
of type Integer
and the following specification definition:
@Spec(path="age", spec=Equal.class)
If non-numeric value is passed with the HTTP request (e.g. ?age=test
), then the result list will be empty. If you want an exception to be thrown instead, use onTypeMismatch
property of the Spec
annotation, i.e:
@Spec(path="age", spec=Equal.class, onTypeMismatch=OnTypeMismatch.EXCEPTION)
This behaviour has changed in version 0.9.0
(exception was the default value in previous ones). The default OnTypeMismatch.EMPTY_RESULT
is useful when using @And
or @Or
and their inner specs refer to fields of different types, e.g.:
@And({
@Spec(path="firstName", params="query", spec=Equal.class),
@Spec(path="customerId", params="query", spec=Equal.class)})
(assuming that firstName
is String
and customerId
is a numeric type)
There is also OnTypeMismatch.IGNORE
type which ignores specification containing mismatched parameter (except spec = In.class
- in this specification only mismatched parameter values are ignored, but other ones which are valid are used to build a Specification).
For example, for the following endpoint:
@RequestMapping(value = "/customers", params = { "id" })
@ResponseBody
public Object findById(
@Spec(path = "id", params = "id", spec = Equal.class, onTypeMismatch = IGNORE) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
- For request with mismatched
id
param (e.g.?id=invalidId
) the whole specification will be ignored and all records from the database (without filtering) will be returned. But for the following endpoint withIn.class
specification type:
@RequestMapping(value = "/customers", params = { "id_in" })
@ResponseBody
public Object findByIdIn(
@Spec(path = "id", params = "id_in", spec = In.class, paramSeparator = ",", onTypeMismatch = IGNORE) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
- For request with params
?id_in=1,2,invalidId
- only valid params will be taken into consideration (invalid params (not the whole specification) will be ignored) - For request with only invalid params
id_in=invalidId1,invalidId2
- an empty result will be returned as there are only invalid parameters (which are ignored).
Although in pure RESTful API this feature should not be needed, it sometimes might be useful to use values from path variables. Path variables are uri fragments resolvable with Spring's @PathVariable
annotation. You can refer to them by using pathVars
property of @Spec
(instead of params
property). For example:
@RequestMapping("/customers/{customerLastName}")
@ResponseBody
public Object findNotDeletedCustomersByFirstName(
@Spec(path = "lastName", pathVars = "customerLastName", spec=Equal.class) Specification<Customer> spec) {
return repository.findAll(spec);
}
This will handle request GET /customers/Simpson
as select c from Customers c where c.lastName = 'Simpson'
.
Basic regular expressions are supported for path variable matching. All patterns supported by Spring AntPathMatcher are supported. For example:
@RequestMapping(value = "/pathVar/customers/{customerId:[0-9]+}")
@ResponseBody
public Object findById(
@Spec(path = "id", pathVars = "customerId", spec = Equal.class) Specification<Customer> spec) {
return customerRepo.findAll(spec);
}
When a path variable is missing then exception is thrown (typically it means that the mapping is incorrect). I.e. by default @Spec
annotation has missingPathVarPolicy
set to EXCEPTION
. If you want to use multiple paths with different pathVars
, then most probably you should set it to IGNORE
. See the example below:
@RequestMapping(path = { "/customers/{customerId}", "/customers-ref/{customerRefCode}" })
@ResponseBody
public Object getCustomerDetails(
@Or({
@Spec(path = "id", pathVars = "customerId", spec = Equal.class, missingPathVarPolicy = MissingPathVarPolicy.IGNORE),
@Spec(path = "refCode", pathVars = "customerRefCode", spec = EqualIgnoreCase.class, missingPathVarPolicy = MissingPathVarPolicy.IGNORE)
}) Specification<Customer> spec){
return customerRepo.findAll(spec);
}
If missingPathVarPolicy was set to EXCEPTION
, then resolving path variables for the method above would not be possible, as it would always contain at least one unresolved path variable.
This is not a best practice for RESTful services but sometimes you may want to filter with request header's value. You
can resolve request headers with Spring's @RequestHeader
annotation. You can refer to them by using headers
property
of @Spec
(instead of params
property). For example:
@RequestMapping(value = "/customers/reqHeaders")
@ResponseBody
public Object findCustomersByGenderAndNickName(
@RequestHeader(value = "customerGender", required = false) String gender,
@RequestHeader(value = "customerNickName", required = false) String nickName,
@And({
@Spec(path = "nickName", headers = "customerNickName", spec = Equal.class),
@Spec(path = "gender", headers = "customerGender", spec = Equal.class)
}) Specification<Customer> spec){
return customerRepo.findAll(spec);
}
This will handle request GET /customers/reqHeaders
as select c from Customers c where c.gender = :gender AND c.nickName = :nickName
.
Also, you can specify value for specification in json request body. It might be useful when you use large number of filters for request because request url is limited in size. You can refer to the specification values by using jsonPaths
property of @Spec
(instead of params
property).
In order to use specification with json request body params, gson
dependency has to be added to the project.
Example maven dependency in project pom file:
<dependency>
<groupId>com.google.code.gson</groupId>
<artifactId>gson</artifactId>
<version>2.10.1</version>
</dependency>
Warning! RequestBody with specification values will be consumed during processing and will not be available for further operations (i.e. ServletInputStream will return no data). If you need to use request body for something else (rather than just building the Specification), you need to use some kind of content caching/wrapping (e.g. in a servlet filter).
For example:
@PostMapping("/customers/find")
public List<Customer> findCustomersByLastNameAndAge(
@And({
@Spec(path = "lastName", jsonPaths = "customerLastName", spec = Equal.class),
@Spec(path = "age", jsonPaths = "customerAge", spec = Equal.class)
}) Specification<Customer> spec) {
return repository.findAll(spec);
}
This will handle request POST /customers/find
with body:
{
"customerLastName": "Simpson",
"customerAge": 18
}
as select c from Customers c where c.lastName = 'Simpson' and c.age = 18
Nested json objects are supported. You should specify full path to node from root element dividing nodes by .
@PostMapping("/customers/find")
public List<Customer> findCustomersByLastNameAndGender(
@And({
@Spec(path = "lastName", jsonPaths = "filters.customer.lastName", spec = Equal.class),
@Spec(path = "gender", jsonPaths = "filters.gender", spec = Equal.class)
}) Specification<Customer> spec) {
return repository.findAll(spec);
}
This will handle request POST /customers/find
with body:
{
"filters": {
"customer": {
"lastName": "Simpson"
},
"gender": "MALE"
}
}
as select c from Customers c where c.lastName = 'Simpson' and c.gender = 'MALE'
For multiple values you can use array (of primitive types only) as result node in json body. For example:
@PostMapping("/customers/find")
public List<Customer> findCustomersByGenderIn(
@Spec(path = "lastName", jsonPaths = "filters.genders", spec = In.class) Specification<Customer> spec) {
return repository.findAll(spec);
}
This will handle request POST /customers/find
with body:
{
"filters": {
"genders": ["MALE", "FEMALE"]
}
}
as select c from Customers c where c.gender in ('MALE', 'FEMALE')
!!!ATTENTION: Json cannot contain array of non-primitive types (array of objects). For example:
@PostMapping("/customers/find")
public List<Customer> findCustomersByLastName(
@Spec(path = "lastName", jsonPaths = "customer.names.firstName", spec = Equal.class) Specification<Customer> spec) {
return repository.findAll(spec);
}
Request POST /customers/find
with the following body is not valid (JsonParseException
will be thrown)
{
"customer":{
"names":[
{
"firstName":"value1"
},
{
"lastName":"value2"
}
]
}
}
Json body parameters are optional in requests, and they can be omitted from the payload. To successfully resolve the parameters, requests must define a Content-type: application/json header and must have a JSON request body, otherwise the parameters will not be recognized and the criteria will not be applied. As most client implementations (e.g. fetch) do not allow GET or DELETE requests with body and Content-type, Json body parameters can be practically used only with POST, PUT and PATCH requests.
Specification argument resolver uses conversion mechanism to convert request string params to types of fields for which specifications have been defined.
Let's consider the following code:
@Controller
public class SampleController {
@Autowired
CustomerRepository customerRepository;
@RequestMapping("/find")
public List<Customer> findByRegistrationDate(
@And({
@Spec(path = "name", params = "name", spec = Equal.class),
@Spec(path = "registrationDate", params = "registrationDate", spec = GreaterThanOrEqual.class)
}) Specification<Customer> spec) {
return customerRepository.findAll(spec);
}
}
@Entity
public class Customer {
String name;
Date registrationDate;
}
When the following request will be sent to the endpoint presented above
GET /find?name=John®istrationDate=2020-06-19
a Specification<Customer>
based on fields name
and registrationDate
will be built.
- The type of the
name
field is aString
type, received parameter value is always aString
type so there is no need of conversion. - The type of the
registartionDate
field isjava.util.Date
type, so theString
will be converted intoDate
using one of available converter.
Specification Argument Resolver contains converters for most common java types.
List of supported conversions:
String -> Enum
String -> boolean
String -> Boolean
String -> integer
String -> Integer
String -> long
String -> Long
String -> float
String -> Float
String -> double
String -> Double
String -> BigDecimal
String -> Date
(default format:yyyy-MM-dd
)String -> Calendar
(default format:yyyy-MM-dd
)String -> LocalDate
(default format:yyyy-MM-dd
)String -> LocalDateTime
(default format:yyyy-MM-dd'T'HH:mm:ss
)String -> OffsetDateTime
(default format:yyyy-MM-dd'T'HH:mm:ss.SSSXXX
)String -> Instant
(default format:yyyy-MM-dd'T'HH:mm:ss.SSSXXX
)String -> Timestamp
(default format:yyyy-MM-dd'T'HH:mm:ss.SSS'Z'
)String -> UUID
To use a custom format for temporal types, add config="custom-format-value"
to @Spec
params.
For example:
@Spec(path="creationDate", spec=LessThan.class, config="dd-MM-yyyy")
In case of missing converter, fallback mechanism will be used if one has been configured otherwise ClassCastException
will be thrown.
If for date-time formats which store also time (LocalDateTime
, OffsetDateTime
, Instant
and Timestamp
) only the date is provided, then time value will be set to the default value - midnight (UTC time for OffsetDateTime
). For example, let us assume that the above specification with the custom config config="dd-MM-yyyy"
corresponds to the LocalDateTime
field in a database. Each argument provided to the specification will be converted to the date with the default time (e.g. 14-12-2022
-> 14-12-2022 00:00
)
The formats of the date in the database (column storing date/time/datetime) and corresponding Java object should be compatible. Also pay attention to the default format for specific date types (if they store date, time or date and time). Inconsistencies in the date formats may lead to confusing or empty results. For example, the database column storing date with time and mapped to Date
Java object cannot be filtered by time until custom config is specified (default config refers only to date yyyy-MM-dd
). It can be achieved by specifying custom format using config
property of @Spec
(e.g. config="yyyy-MM-dd'T'HH:mm:ss.SSS'Z'"
).
The converter includes a fallback mechanism based on Spring ConversionService
, which is invoked when required conversion is not supported by any default converter. If the ConversionService
supports required conversion it will be performed, otherwise a ClassCastException
will be thrown.
To add specification argument resolver support for custom conversion:
- Add required converter to ConversionService
- Configure fallback mechanism by passing ConversionService to SpecificationArgumentResolver constructor.
For example:
@Configuration
@EnableJpaRepositories
public class MyConfig implements WebMvcConfigurer {
@Autowired
ConversionService conversionService;
@Override
public void addArgumentResolvers(List<HandlerMethodArgumentResolver> argumentResolvers) {
argumentResolvers.add(new SpecificationArgumentResolver(conversionService));
}
@Override
public void addFormatters(FormatterRegistry registry) {
registry.addConverter(new StringToAddressConverter());
}
public static class StringToAddressConverter implements Converter<String, Address> {
@Override
public Address convert(String rawAddress) {
Address address = new Address();
address.setStreet(rawAddress);
return address;
}
}
...
}
Locale is important when using case-insensitive specifications such as EqualIgnoreCase
. For example, in Turkish, the uppercase form of i
is İ
(U+0130, not ASCII) and not I
(U+0049) as in English.
Specification-arg-resolver uses system-default (Locale.getDefault()
) locale if no explicit configuration is provided. You can configure custom default locale globally, during argument resolver registration:
@Override
public void addArgumentResolvers(List<HandlerMethodArgumentResolver> argumentResolvers) {
argumentResolvers.add(new SpecificationArgumentResolver(new Locale("pl", "PL"))); // pl_PL will be used as the default locale
}
You can also configure it for each individual specfication definition via @Spec.config
(this overrides the global default mentioned above):
@Spec(path = "name", spec = EqualIgnoreCase.class, config = "tr_TR")
Support for SpEL expression and property placeholders can be enabled in following way:
- Configure
SpecificationArgumentResolver
by passing AbstractApplicationContext in constructor - Set attribute
valueInSpEL
value totrue
Configuration example:
@Autowired
AbstractApplicationContext applicationContext;
@Override
public void addArgumentResolvers(List<HandlerMethodArgumentResolver> argumentResolvers) {
argumentResolvers.add(new SpecificationArgumentResolver(applicationContext));
}
SpEL expressions can be applied to @Spec
constVal
, defaultVal
and params
. The first two are described in more detail in corresponding sections above. SpEL support for params
can be enabled via @Spec.paramsInSpEL
. It may be useful in rare cases when you want to differentiate HTTP parameter name based on the application configuration or other contextual attributes.
Right now specification argument resolver supports only one library -> Springdoc-openapi
.
There are two steps in order to enable support for Springdoc-openapi
library:
-
Add dependency from
Springdoc-openapi
:- For
specification-arg-resolver 3.0.0
and newer, please usespringdoc-openapi-starter-common
(tested with2.0.2
version):<dependency> <groupId>org.springdoc</groupId> <artifactId>springdoc-openapi-starter-common</artifactId> </dependency>
- For versions older than
specification-arg-resolver 3.0.0
, please usespringdoc-openapi-common
(tested with1.6.13
version):<dependency> <groupId>org.springdoc</groupId> <artifactId>springdoc-openapi-common</artifactId> </dependency>
- For
-
Create
@Bean
of typeSpecificationArgResolverSpringdocOperationCustomizer
in your app configuration:
@Bean
public SpecificationArgResolverSpringdocOperationCustomizer specificationArgResolverSpringdocOperationCustomizer() {
return new SpecificationArgResolverSpringdocOperationCustomizer();
}
Specification argument resolver supports spring cache. Equals and HashCode contract is satisfied for generated specifications.
Specification argument resolver supports creating specifications apart from web layer.
To build specification outside the web-layer the SpecificationBuilder
should be used:
-
Let's assume the following specification:
@Join(path = "orders", alias = "o") @Spec(paths = "o.itemName", params = "orderItem", spec=Like.class) public interface CustomerByOrdersSpec implements Specification<Customer> { }
- To create specifications outside the web layer, you can use the specification builder as follows:
Specification<Customer> spec = SpecificationBuilder.specification(CustomerByOrdersSpec.class) // good candidate for static import .withParam("orderItem", "Pizza") .build();
- It is recommended to use builder methods that corresponding to the type of argument type passed to specification interface, e.g.:
- For:
you should use@Spec(paths = "o.itemName", params = "orderItem", spec=Like.class)
withParam(<argName>, <values...>)
method. Each argument type (param, header, path variable) has its own corresponding builder method:params = <args>
=>withParam(<argName>, <values...>)
, single param argument can provide multiple valuespathVars = <args>
=>withPathVar(<argName>, <value>)
, single pathVar argument can provide single valueheaders = <args>
=>withHeader(<argName>, <value>)
, single header argument can provide single value
The builder exposes a method
withArg(<argName>, <values...>)
which allows defining a fallback value. It is recommended to use it unless you really know what you are doing. - To create specifications outside the web layer, you can use the specification builder as follows:
This project has been maintained since 2014. A lot has changed in Java and Spring during that period and the most recent versions might not be compatible with older JDK and/or Spring. In the table below you can find the summary of version compatibility:
specification-arg-resolver version | JDK requirements | Spring requirements |
---|---|---|
v3.X |
17 or higher |
Compiled and tested against Spring Boot 3.0.0 |
v2.X |
1.8 or higher |
Compiled and tested against Spring Boot 2.7.7 |
v1.1.1 (or older) |
1.7 or higher |
Compiled and tested against Spring Boot 1.x ; confirmed to work with Spring boot 2.x |
As far as the features supported in each version, please check the CHANGELOG.md
Specification argument resolver is available in the Maven Central:
<dependency>
<groupId>net.kaczmarzyk</groupId>
<artifactId>specification-arg-resolver</artifactId>
<version>3.1.0</version>
</dependency>
If a new version is not yet available in the central repository (or you need a SNAPSHOT version), you can grab it from my private repo:
<repository>
<id>kaczmarzyk.net</id>
<url>http://repo.kaczmarzyk.net</url>
</repository>