I'm off to bed, I just created <https://github.com...
# prisma-whats-newc
cj
08/30/2017, 7:16 AMI'm off to bed, I just created https://github.com/cj/jsgql, thought I'd link it here, as it might be useful to some
👍 2
d
Daniel K.
08/30/2017, 7:36 AMlooks interesting, but I cannot think of any use case for it 🙂
d
dankent
08/30/2017, 8:58 AMI think it could be useful when programatically generating graphql. Instead of doing a big string concatenation exercise to make the string to pass to gql, you would build an object and then use this to generate the graphql AST from it
d
Daniel K.
08/30/2017, 9:08 AMwell i think that string concat is still easier 🙂
d
dankent
08/30/2017, 11:29 AMeach to their own but I'm definitely going to give this a try next time I need to generate some dynamic, highly nested graphql - I suspect dealing with objects will be less messy and easier to maintain than strings
c
cj
08/30/2017, 12:56 PM@dankent exactly right. I got tired of creating lots of
.gqlcj
08/30/2017, 12:58 PM@Daniel K. string concat is messy if you are doing it all inline, it distracts from your actual code and can make it pretty messy if you're doing it all the time. This does it that for you behind the scenes, so all you deal with is a nice js object 🙂
d
Daniel K.
08/30/2017, 1:06 PMyea well, I find an AST too bloated for it to be able built it easily, it would need another wrapper that could generate with same useful interface, building whole object could be messy same way
d
dankent
08/30/2017, 1:07 PM@cj something that would be useful would be if this would support multiple queries/mutations within a single outer query
dankent
08/30/2017, 1:08 PMThat would mean having a methods array rather than than single method property and enabling the alias for each one to be set
dankent
08/30/2017, 1:08 PMc
cj
08/30/2017, 1:08 PM@dankent you'd just do
[fooGQL, barGQL].map(gql => jsgql(gql))cj
08/30/2017, 1:09 PM@Daniel K. I'd love to see an example
d
dankent
08/30/2017, 1:09 PMwouldn't that make separate graphql AST's - I'm talkiing about multiple operations within a single AST
dankent
08/30/2017, 1:10 PMSituations like the one in the link above are often where I end up do dynamic graphql generation, to get things like:
dankent
08/30/2017, 1:10 PMmutation {
charmander: createPokemon(name: "Charmander", type: FIRE) {
type
}
bulbasaur: createPokemon(name: "Bulbasaur", type: GRASS) {
type
}
}
c
cj
08/30/2017, 1:11 PMI can make it do that if you parse an Array to jsgql, the reason I haven't is Apollo client batching does that fo you and more intelligently, which graph.cool supports
cj
08/30/2017, 1:12 PMmeaning you can call apollo.query in multiple places in your app concurrently and it will batch them all together in a single call 🙂
d
dankent
08/30/2017, 1:13 PMUsing automated batching doesn't provide guarantees about mutation execution order. Putting multiple aliased mutations in one graphql mutation element guarantees that the mutations will be executed in series
dankent
08/30/2017, 1:14 PMWith automated batching, if my mutations end up being sent in separate requests, graphql will execute them in parallel
c
cj
08/30/2017, 1:15 PMI can add that in if you like, I just hadn't had a use case for it, if I needed something to come after something, I would just nest they query fields like
id customers { id }cj
08/30/2017, 1:16 PM@dankent do you have an example use case for that? I'd love to see one as I haven't come across it yet
cj
08/30/2017, 1:16 PMI'll add the acceptance of an array today though, which will give that output you pasted
d
dankent
08/30/2017, 1:18 PMTwo main situations where you want to guarantee serial execution of mutations:
dankent
08/30/2017, 1:19 PM1) You are doing something in the request pipeline that takes time (e.g. an external call) and you want to ensure it completes before you do the next mutation
dankent
08/30/2017, 1:20 PM2) You want subsequent mutations to be cancelled if one before it in the list fails
dankent
08/30/2017, 1:20 PMaliased mutations within a single mutation give you both of these behaviours
dankent
08/30/2017, 1:21 PMApollo does batching of queries but does not (as far as I know) do any automated batching of mutations
dankent
08/30/2017, 1:21 PM(which is a good thing - if it did, failure of one mutation in the batch could cause other unrelated mutations to fail)
c
cj
08/30/2017, 1:25 PMfor me, I would never use that use case, because now you can't rely on the automatic grapqhql docs, as you now have to specifically tell people, "HEY when you do this mutation don't for get to run this one first!"
cj
08/30/2017, 1:26 PMI would handle that logic behind the scene so it's just a nice single simple graphql mutation
d
dankent
08/30/2017, 1:31 PMThe cases where I have had to do this, it's obvious that the subsequent mutations don't make sense without the first one happening before them. I will be reworking at least some of them as schema extensions to provide a single call but I expect that aliased mutations will still be a useful tool from time to time.
n
nilan
08/30/2017, 1:49 PM@cj @dankent would love for you to chime in here: https://www.graph.cool/forum/t/query-batching-in-graphcool-vs-apollo/409?u=nilan 🙂
nilan
08/30/2017, 1:50 PMnotably, Apollo Query Batching is different to "using the same field multiple times in the same GraphQL document-batching"
nilan
08/30/2017, 1:50 PM@cj feel free to share your project here as well: https://www.graph.cool/forum/c/show-tell 🙌
c
cj
08/30/2017, 2:15 PM@nilan morning. will do! I didn't know about that part of the forum 🙂
🎊 1