When Facebook first previewed their upcoming graph data layer, I found the concepts compelling, and the query language promising. But when it finally arrived—buried largely within the bizarre abstractions and verbose oddities of relay (getFatQuery?)—the tooling fell well short of the promise. I must confess to briefly adopting a competitor that is now at best an also-ran in this space—Netflix’s falcor—simply due to the cleaner and less weighted-down API of the latter.

Things began to change when Apollo stepped in and finally began to build tools for GraphQL which proved its generalizable utility outside of Relay. Within a short time spent adapting a large Meteor application to use GraphQL for all its fetching and mutations, I was hooked on the concept and began to use it for a number of projects in different domains, even creating a few small wrappers to mimick the structure of apollo-tools within the python implementation of GraphQL of that time. I have used Apollo’s various implementations, secondary tools (schema stitching, local state management, etc), and other features like subscriptions many times over the past years.

In the context of static-site generation, I found Gatsby’s adoption of GraphQL for its data layer to be one of the features that has made it the ideal tool for static sites. At various times, I have contributed to its codebase in this area.

In my university position, I developed internal tools for using GraphQL as a complete platform for describing and generating deeply nested (meaning, “pivoted” but with finer control) descriptive statistics, permitting my front-end code to declaratively describe the shape of any deep hierarchy of filters, grouping, or statistics, which are then converted immediately into a GraphQL query, sent to the server, and populated into tables (see the basic tests for retabulate-graphql). More information on the initial motivation for that strategy is available under SAS, and it also forms a major part of my common reporting strategy.

At Slite, our application’s data layer was also served via GraphQL, so I used it while developing a variety of front-end features.