This is my first te(a)ch note on GraphQL. I had hit a couple road blocks in a few of days of my hands on journey with it. Unlike good-old-days when books were the primary source of learning that had everything documented, there is no single place to find all details these days.
Environment: Java 21, Spring Boot 3.2.3, PostgreSQL 15.3, maven 3.9.6 on macOS Catalina 10.15.7
Extended or Custom Scalar types
GraphQL specifies very limited set of well-defined built-in scalar data types (primitive data types): Int, Float, String, Boolean and ID. GraphQL systems must support these as described in the specification. Everything else is an extended or custom scalar data type.
That, obviously is a very limited set supported. All other data types need custom scalar implementations which basically require coercing values at run-time and converting those to Java run-time representation. Luckily, the Java ecosystem is so huge that you almost don't need to break the ground in doing so. You will always find few open-source libraries that have tackled it already for you. GraphQL Java Scalars is one such in this Java GraphQL for extended scalar data types.
The primitive data type set supported is just not enough. You at least need support for few other data types used in any Java application like: Long, UUID, DateTime etc. They all need special considerations in your application's GraphQL schema. The DateTime takes a very special seat. In fact, anything around Dates in Java always scares me. To humans, Date and Time are the most obvious types in day-to-day life, but non in Software systems. Date is the most abused type than any other data type. Just recollect how many billions of dollars of money was wasted on this one data type in 1998 and 1999 around the globe. After 23 years of learning the mistake, the Date is still not dealt easily; it is still a complex data type to deal with ;).
To use custom scalar types other than that limited primitive set, you have to look for writing code that handles serialization, parsing and literal parsing for each additional data type. The graphql-java-extended-scalars library provides implementation for many other data types.
With a maven dependency added for this library, all you need to do is to register a scalar data type with RuntimeWiringConfigurer as described in the README. If you need to register multiple types, it's a builder, so you can just chain those like:
@Configuration
@Slf4j
public class GraphQLConfiguration {
/**
* Custom scalar support for UUID, Long, and DateTime. Registers extended scalar types used GraphQL query schema.
*/
@Bean
public RuntimeWiringConfigurer runtimeWiringConfigurer() {
log.info("Registering extended GraphQL scalar types for UUID, Long, DateTime...");
return wiringBuilder -> wiringBuilder.scalar(ExtendedScalars.UUID)
.scalar(ExtendedScalars.GraphQLLong)
.scalar(ExtendedScalars.DateTime);
}
}
"Extended scalar types"
scalar UUID
@specifiedBy(url: "https://tools.ietf.org/html/rfc4122")
scalar Long
@specifiedBy(url: "https://ibm.github.io/graphql-specs/custom-scalars/long.html")
scalar DateTime
@specifiedBy(url: "https://scalars.graphql.org/andimarek/date-time.html")
...
Note that the extended scalar type for Long is named as GraphQLLong by this library. But, you should use Long in your your schema when you specify it as shown above. The directive @sprifiedBy is recommended to be used by GraphQL specification and is also a good practice to follow. Never ignore good practices ;)
Gotcha
Java JPA - Instant vs. OffsetDateTime
If you are dealing with DateTime, make sure that whatever the Java type used in your code, it complies with GraphQL specification that requires date time offset.
I initially used Instant type in my JPA BaseEntity class for two properties: createdOn and updatedOn that are mapped by Hibernate provided @CreationTimestamp and @UpdateTimestamp mapped to PostgreSQL column type TIMESTAMP WITH TIME ZONE. I switched to OffsetDateTime type because Instant is not supported and will never be by this library due to it not complying with the specification for DateTime. Java's Instant, Date and LocalDateTime do not include offset.
OffsetDateTime is an immutable representation of a date-time with an offset. This class stores all date and time fields, to a precision of nanoseconds, as well as the offset from UTC/Greenwich.
TIP
PostgreSQL offers two date time types: timestamp, timestamptz (is abbreviation of timestamp with time zone).
The following query results tell the date time story on this day light savings day of this year (Sun Mar 10, 2024). I ran it on my local PostgreSQL 15.3 running in Docker container.
-- Ran the query on Mar 10, 2024 day light savings day at EST 5:13:13 PM, EDT: 17:13:13
select version(); -- PostgreSQL 15.3 (Debian 15.3-1.pgdg120+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 12.2.0-14) 12.2.0, 64-bit
show time zone; -- UTC
SELECT now(), -- 2024-03-10 21:13:13.956877 +00:00 (timestamp with time zone UTC)
now() AT TIME ZONE 'EST' AS est, -- 2024-03-10 16:13:13.956877 (??)
now() AT TIME ZONE 'EDT' AS edt, -- 2024-03-10 17:13:13.956877 (right)
now() AT TIME ZONE 'CST' AS cst, -- 2024-03-10 15:13:13.956877 (??)
now() AT TIME ZONE 'CDT' AS cdt, -- 2024-03-10 16:13:13.956877 (right)
now()::timestamp AT TIME ZONE 'EDT' AS timestamp_without_tz, -- 2024-03-11 01:13:13.956877 +00:00 (wrong)
now()::timestamptz AT TIME ZONE 'EDT' AS timestamptz; -- 2024-03-10 17:13:13.956877 (right)
Here is the DbFiddle playground to play with the above query.
That's it in this te(a)ch note, more might come in as I walk forward along this GraphQL path.
Posts
No comments:
Post a Comment