nanograph

An experimental graph client investigating how much logic can be included in TypeScript types to reduce the overall library size while supporting fluent query patterns, typing support, and intellisense.

The goal is not 100% perfect coverage, but how close can we get with minimal code. The solution uses a Proxy to interpret the calls and dynamically build and execute the requests. The total minimized, uncompressed package size is ~50k.

If we remove the ability to call actions/functions we can reduce the size to < 4k which still includes paging, batching, get, update, add, select, filter, expand, and full fluent developer experience support.

Considering most data access uses the core patterns of getting/updating collections and getting/updating individual items a similar approach could have significant impact on page load times and performance for clients, especially mobile and markets that do not have highspeed internet. It also introduces the possibility with more time to use dynamic imports to initially provide the core capabilities in a very small package and bring in more as needed.

Sending less traffic helps reduce COGS at server and client as well as reducing memory pressure by loading a smaller package.

There are many sample queries shown in main.ts including batching, paging using async iteration, basic gets, and adds.

setup

Add a node-auth-settings.ts file into ./src with the form:

const privateKey = `-----BEGIN RSA PRIVATE KEY-----
-----END RSA PRIVATE KEY-----
`;

export const configuration = {
    auth: {
        authority: "https://login.microsoftonline.com/{tenant}/",
        clientCertificate: {
            thumbprint: "{thumbprint}",
            privateKey,
        },
        clientId: "{client id}",
    }
}

export const scopes = ["https://graph.microsoft.com/.default"];

Run npm install
Update examples in main.ts then hit F5 to run

patrick-rodgers/fhl-2023-fall

nanograph

setup