add options to load() to resolve recursiveDelete() issue

[bourgeoa]

1. this.store.fetcher.load(doc) with no options reuse the existing document in store.
   This is not what we expect. We want load a new document.
   options.force = true reload the document but adds new triples
   we neee to remove the document from store in that case

2. force = true is used in aclDocUrl() and getContainerMembers().
   @michielbdejong Should it be a default for load() ?

3. solid spec implies that delete is atomic (delete ressource and auxiliary one in the same operation)
   I remove the acl delete in recursiveDelete()

[edwardsph]

LGTM

[edwardsph]

I'm quite new to Typescript so may have misunderstood but if someone passed a NamedNode array or a string to load() will it work to pass that to removeDocument as a NamedNode?

[bourgeoa]

Yes it will work if you don't use force = true. If you use force = true doc must be a Named node.
I took this solution because it was used elsewhere in the code, it was simple, it was compatible with the used cases and offer a plus possibility to load(). In an other way there is no regression

If you don't use force = true then load can receive NamedNode, NamedNode Array or String like before.

[bourgeoa]

@edwardsph I made the changes. I don't know anything in typescript. My initial idea was wrong. Load now can pass a NamedNode, NamedNode Array or a string to load() and pass that to removeDocument() as namedNode.

Tested locally with all 3 types.

[edwardsph]

LGTM

[timbl]

"his.store.fetcher.load(doc) with no options reuse the existing document in store.
This is not what we expect. We want load a new document." -
No no no no no. The semantics of load is "loadIfNotAlreadyLoaded" That is really important. It is the porose version of a function "nowOrWhenFetched" which would call back immediately oif the thing had already been loaded into the store. The caching of documents in te quad store makes everything VERY MUCH faster. And the solid-ui code calls x.load() many times just to make sure it is loaded before following links from x. It is essentilal that x.load returns immediately if tehe document is already noted. We clearly have to make this clear in the documentation

[timbl]

If you want to reload something already loaded you have refresh https://github.com/linkeddata/rdflib.js/blob/master/src/fetcher.ts#L1732
and refershIfExpired which will check the cache against the expiry date on the resource.

There is also fetcher.unload. Do not use store.removeDocument as it will leave the cache metadata wrong.

[bourgeoa]

@timbl Thanks.
As I understand :

• solid-logic offers helper functions to solid-ui as the main purpose as you said in solidos chat
• these helper functions are mainly working on the store in cache

And the solid-ui code calls x.load() many times just to make sure it is loaded before following links from x. It is essentilal that x.load returns immediately if tehe document is already noted. We clearly have to make this clear in the documentation

The proposed patch do not change the actual use of x.load() in solid-ui

Shall there be possibility to refresh a document in the store if expired ? That opens the use case of solid-logic like in test-suite.
and it improves the cache use :

1. should refresh be at the load() level or only in the functions using load() : I prefer load() this makes it global
2. should that be implicit or with explicit options : for performance I think default should not refresh

Do not use store.removeDocument as it will leave the cache metadata wrong.

refreshIfExpired() seems an interesting medium term. But how can I know that an rdflib function is unsafe ?

[bourgeoa]

Finally modified only getContainerMembers(). Recursive delete must be sure to act on the latest situation and not from a cache that may not be up to date.

[timea-solid]

@bourgeoa is this still accurate? Shall we merge it?

[bourgeoa]

For me it was good to merge.
Recursive delete it used in the solid/test-suite
I see some conflicts now

[timea-solid]

Because this PR was hard to merge due to the refactoring of code, I created a new PR with just the prominent changes of this PR: https://github.com/solid/solid-logic/tree/loadImprovement

@bourgeoa please check it and close and merge these PRs if ok.
Tests are failing. Those need to be checked.