What is the use case for returning the cursor of each individual node? about graphql-relay-js

[dschafer]

Note that while Relay requires edges for its pagination, it doesn't have an opinion on whether nodes is exposed. So if you find nodes convenient for browsing APIs (I certainly do), I'd definitely add it in! If you do so, Relay has a check in place so that if you query for nodes instead of edges, it will warn and suggest querying for edges instead.

[yuzhi]

We fetch cursors on every edge so that if you have a component that asks for first:10 and have it cached and later another component asks for first:5, we can generated the correct pageInfo based on the cursors we have fetched from first:10 for first:5 without doing a server round trip.
Another case is, if some of the edges are deleted through mutations, we still have the cursors form the other edges to fetch for additional edges.

[dschafer]

CC @yuzhi who can describe what the Relay store does with the cursors on each edge, and hence why it uses edges rather than nodes.

Internally, we expose nodes for exactly the reason you describe: if you don't need edge data or the cursor, it's quite convenient to skip straight to the nodes; that's definitely a useful pattern.

[luisobo]

Yeah, that's what @leebyron suggested on graphql/graphql-spec#4.

Can't what to hear the purpose of that. I'm guessing something related to nodes being evicted out of a local cache and refetching parts of a connection?

[luisobo]

Interesting thanks!

/cc @modocache @akatz

On Tuesday, September 22, 2015, Yuzhi [email protected] wrote:

We fetch cursors on every edge so that if you have a component that asks
for first:10 and have it cached and later another component asks for
first:5, we can generated the correct pageInfo based on the cursors we
have fetched from first:10 for first:5 without doing a server round trip.
Another case is, if some of the edges are deleted through mutations, we
still have the cursors form the other edges to fetch for additional edges.

—
Reply to this email directly or view it on GitHub
#27 (comment)
.