Experimental note body serialization #1
Conversation
mock out fetch enough to get tests working and avoid calling my pod
don't hardcode properties, just pull out everything we can find
developing from the testfile made sense for a while, but this is much better
There was a problem hiding this comment.
RDF:list seems right to me too. Looks great! I think this is still WIP based on our chats, though most of the library is still WIP until we get it into Garden. Feel free to merge if you want, or lemme know if you do another pass and want me to take another look.
package.json
Outdated
| @@ -44,6 +44,7 @@ | |||
| } | |||
| ], | |||
| "devDependencies": { | |||
| "@ontologies/core": "^2.0.3", | |||
There was a problem hiding this comment.
Why dev deps for this one but not the others?
There was a problem hiding this comment.
oo I think this is a good catch, thx!
src/note.ts
Outdated
| import {arrayToThings, thingsToArray} from "./collections" | ||
|
|
||
| const noteNSUrl = "https://mysilio.garden/ontologies/note#" | ||
| const noteNS = createNS(noteNSUrl) |
There was a problem hiding this comment.
I use @rdf-esm/namespace for this:
import namespace from '@rdf-esm/namespace
See src/vocab for example usage.
There was a problem hiding this comment.
Doesn't matter much which one we use, I don't think, because they both build NamedNodes. I think I even came across the @ontologies one, might be by the same guy?
| for (const pred in thing.predicates) { | ||
| const [, key] = pred.split(noteNSUrl) | ||
| if (key) { | ||
| if (key === 'children') { |
There was a problem hiding this comment.
Why not instead test to see if pred === childrenPred.value or pred === noteNS('children').value?
To me that reads a little cleaner than splitting on the predicate namespace.
| const obj: any = {} | ||
| for (const pred in thing.predicates) { | ||
| const [, key] = pred.split(noteNSUrl) | ||
| if (key) { |
There was a problem hiding this comment.
Just noticed the if (key) -- when would key not exist? If the predicate isn't in our NS? That might explain why you match on key below instead of the predicate.
There was a problem hiding this comment.
oh yep! this is expecting that there might be other predicates on the thing and trying to only deal with things in our NS. I'm really not sure about this particular part of the approach - at the very least I need to double check that my namespaces make sense before shipping this, will take another look at this tomorrow!
src/note.ts
Outdated
| const thing = createThing({ name: (o.id ? `${o.id}` : `el-${path.join("-")}`) }) | ||
| const otherThings = Object.keys(o).reduce((m: Thing[], k: string) => { | ||
| const [mThing, ...mRest] = m; | ||
| const [thing, ...restOfThings] = addKeyValToThing(mThing, noteNS(k).value, o[k], path) |
There was a problem hiding this comment.
It looks like all the places you use this in the downstream function are Inrupt fns. Those all actually take NamedNodes as well as URLStrings for the predicates, so you don't have to do noteNS(k).value here, you could just do noteNS(k) if you wanted.
There was a problem hiding this comment.
hm, I tried this and got:
looks like addUrl doesn't accept a NamedNode?
move vocab-related stuff into vocab.ts and move away from Ontola (ie, @ontologies) libraries there's a subtle incompatibility between the `@ontologies` version of `NamedNode` and the "standard" (ie, rdfjs-base https://github.com/rdfjs-base) versions - the `@ontologies` version doesn't require an `equals` method, so trying to pass `NamedNodes` from that library to the Inrupt libraries doesn't work. Trying to switch would be a big lift for not much work and potentially make us less compatible with `solid-client-js` so I opted to use `@rdfjs/namespace`. Unfortunately the most recent versions of the `@rdfjs` libraries have been converted to ESM modules and are not transpiled to an older version of JS like most NPM libraries, which makes Jest choke at the moment, so I'm following the lead of `solid-client-js` and requiring an older version of this lib. The RDF ecosystem is a smoking wasteland of incompatibility and pain! A real shame because RDF is awesome!
I wanted to use @size-limit/preset-small-lib but there's some weird dependency problem and I don't have the spoons for another trek through npm hell - for now just bump the size limit above the 268k/229k filesizes we have now to get the build working
hoping this fixes the weird build issues I'm seeing
this was originally in garden, moving here to preserve, not ready to merge
use rollup to build the various types of files we need
* create spaces file if it doesn't exist * add MY.Garden.Item type to items
|
Close this PR now that |
Add functions for serializing a note body as RDF:
Use an
rdf:ListCollection as the format - we'd like to minimize changes during save, and this allows us to keep content subject URLs consistent by adding anidparameter to individual elements in the Slate JSON (which Plate does by default for some elements). Therdf:Listformat is better for this thanrdf:Seqbecauserdf:Seqrequires predicate names likerdf:_1, rdf:_2to indicate position in the list, which would cause cascading updates for things later in a content array structure.The functions in
collections.tsare totally abstract and I'm interested in porting them tosolid-clientin the future per inrupt/solid-client-js#1648