Being short-sighted
It’s a mistake that can lead to serious complications later on. Thinking about your schema try being as predictive as possible in terms what can be required later on when your project grows. Even if you think that initial odds of this happening seems very low, take a while to plan through possible future elements like additional fields or output/input object types. If there is even a tiny change that you will need them in the future, having them thought through will prevent your project from breaking changes later on.
Remember: A schema is a most important part of any GraphQL project so make sure to treat it with all due respect!
Inconsistency in naming
It's really important to keep naming conventions consistent. There are some common practices you shouldn't ignore, like:
- using camelCase for fields
- using PascalCase for the names of types
- using all capital letters for ENUM values
Flattening objects in queries
There are a lot of mistakes you can make when "speaking" GraphQL. Not only mistakes that might prevent your project form working properly, but also that they won't use GraphQL possibilities to the fullest. A good example is a type we have created above.
The leageId
is a problem here as this will be seen as an inconsistency. When querying for a team’s league you need to perform two database calls:
- to get the team with the
leagueId
- to pass the
leagueId
to another query
This is very negligent as GraphQL it's all about reducing the number of database calls. The GraphQL preferable way is nesting the output types which makes it easier to call your database with a single request and batch your data:
Not using graphical interfaces
GraphQL is a very modern "language" so there are a lot of amazing tools, plugins or editors that can speed up the schema design process so why not using them? If you are starting your journey with GraphQL it is a great idea to design your first schema with a graphical interface. This will reduce the number of mistakes, speed up your work and learning curve of GraphQL by a lot!
Not using GraphQL
If you are still focusing on REST you are making a canonical mistake. As none of them is perfect and both have pros & cons, it's really wise to know them both and then make a rational decision which would suit your project better.
Top comments (9)
Nice write up! I don't have any GraphQL experience, but I'm trying to get into it. I've got one question about the paragraph:
As far as I understand it, DataLoader's batching works on the event-loop level. When it comes to SQL databases it'd still be the responsibility of the backend that fetches the data from the database to actually provide the right (optimized) query, right? I mean, a batch could still result in several SQL SELECTs, if a query with a, e.g., WHERE clause, isn't available.
In my opinion dataLoader is more abstract tool which can be used within specific patterns. It doesn't replace SQL queries of course, but you might be confused by a lot of SQL examples inside README. Using it with resolvers and GraphLQ is up to you. You can reduce the number of calls if you want to query just exact fields.
+1 - I have same question. API returns the results all at once, but database get queried in each loop, instead of joins that are possible in most SQL data sources.
For graphql, what i need is automatic fragments, typings, queries, mutations generation. Those are selling points of graphql, so if you have to manually type
graphql query
, you're doing it non-optimally.Yes, yes drop REST at all... What a fraud. Would you implement a single SP to fetch all data from a database? Why the hell one would think doing that with http requests world work nicely?
Calm down :) What I meant is that you should stop being blindfolded and open your eyes on GraphQL.
You are right, @Tomek :) I've just read too many hype articles praising GraphQL as a silver bullet and I took it all on yours ... my bad.
Thanks very useful!
Thanks, that was useful!