How reactivity works
Thanks for the feedback! So whenever a table is
.query'd, we keep track of the read ranges -- what index was used and what range within that index4 Replies
So if a given query function read records from two different tables at two different indexes, we "track" the fact those records were used to produce the final return value of that function
then, a mutation comes by that
insert, patch/update, or delete some record(s)
for each of those records, we run a kind of "intersection" to see if it is within a read range of any subscribed queries
if so, all of those queries are invalidated and re-run to produce a potentially new result. and that new result is pushed over the web socket to the convex client library
this is a somewhat simplified version of how it works
we do have a blog post underway called "How Convex Works" that goes into great detail about all thisThat makes sense, very cool - i'm looking forward to read the blog post. It's very interesting to me. Thanks Jamie!
if a query does something like a filter and then a sum over a table, then returns the first few records from the result, will the query be re-run when items are added to the db? how does it work for those cases
it will re-run, yes. it will basically always re-run when it should
the reason is when you add items to the database, they are inserted into the range in an index the query was based on. so the query knows the result might have changed, and so it is re-run