Re: [w3c/clipboard-apis] Defer OS clipboard read from clipboard.read() to ClipboardItem.getType() (PR #248)

@EdgarChen commented on this pull request.



> @@ -918,6 +934,60 @@ url: https://storage.spec.whatwg.org/#obtain-a-storage-key-for-non-storage-purpo
 
     1. If |representation|'s [=representation/MIME type=] is |mimeType| and |representation|'s [=representation/isCustom=] is |isCustom|, then:
 
+     1. If [=this=]'s [=ClipboardItem/clipboard change count at read=] is not null, and the current [=clipboard change count=] is not equal to [=this=]'s [=ClipboardItem/clipboard change count at read=], then [=reject=] |p| with an {{"InvalidStateError"}} {{DOMException}} in |realm|, and return |p|.
+
+      Note: This ensures that stale data is never returned. If the [=system clipboard=] contents have changed since {{Clipboard/read()}} was called, this MUST fail rather than returning data that does not correspond to the current [=system clipboard=] state.
+
+      Note: A null [=ClipboardItem/clipboard change count at read=] indicates that this {{ClipboardItem}} was constructed directly by the author (e.g., via {{ClipboardItem/ClipboardItem(items, options)}}), not obtained from {{Clipboard/read()}}. In that case, no stale-data check is needed because the data does not originate from the [=system clipboard=].
+
+     1. If [=this=]'s [=ClipboardItem/clipboard change count at read=] is not null, then:
+
+      Note: This branch covers {{ClipboardItem}}s obtained from {{Clipboard/read()}}, for which the OS clipboard read is deferred to this point. To ensure the read happens at most once per type — even if {{ClipboardItem/getType()}} is called multiple times concurrently for the same MIME type — the resulting {{Promise}} is cached in [=this=]'s [=ClipboardItem/representations with resolvers=] map, keyed by a normalized type string so that `"text/html"` and `"web text/html"` are tracked independently while differently-cased spellings of the same type (e.g., `"text/HTML"` and `"text/html"`) collapse to a single entry.
+
+      1. Let |key| be |mimeType|'s [=MIME type/essence=]. If |isCustom| is true, prefix |key| with `"web "`.
+
+      1. If [=this=]'s [=ClipboardItem/representations with resolvers=][|key|] [=map/exists=], then return [=this=]'s [=ClipboardItem/representations with resolvers=][|key|].

So if the `ClipboardItem` is created by `read()`, two `getType()` calls on it can return same promise instance. And if the `ClipboardItem` is constructed from script, `getType()` return different promise instance. All browsers currently always return different promise in either case. Is this intentional behavior changes and differences? 


-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3c/clipboard-apis/pull/248#pullrequestreview-4544850370
You are receiving this because you are subscribed to this thread.

Message ID: <w3c/clipboard-apis/pull/248/review/4544850370@github.com>