… unlikely due to low-level api)

datomic-m setup is broken with this addition, commented out from
startup, fix once needed

Local edits are stored as if on a branch. Clicking commit merges changes
back into the global transactor. Clicking discard throws the local
changes away.

With multiple editors double-clicking on 1 todo item would open the item
for editing in all editors. This fix opens the item for editing only in
the 1 editor we clicked it.

The multi-editor setup was broken anyway because
- focus could end up in the wrong editor
- edits weren't propagating to other editors

We expect to have users of ui4/button and don't want to force inline
styles on them.

Our new Datomic demos reading off the tx queue create a singleton reader
of the queue since multiple readers steal values off the queue (it is
not pub/sub). This singleton is a separate thread and we need to make
sure it doesn't block the JVM from shutting down (i.e. make them
daemon).

The TodoMVC demos now use daemon threads. The advanced todolist demos
are removed from the requires since

- they need updating to use the daemon thread pattern
- are not linked to anyway since they are in early prototype stage

… in callbacks

instead of (e/offload #(db-query a b c))
now write  (e/offload-latest db-query a b c)

differences:
- takes *electric* varargs instead of clojure thunk
- doesn't return intermediate Pendings. This fixes over-rendering and
  cycles in callbacks because work-skipping will correctly kick in.

Might not be the final solution to this problem space, but a fine
solution for now as it fixes all user concerns and doesn't need
e/snapshot.

document weird behavior in client/server transfer where we seem to see
an old value. Not in all cases though, 3 out of 4 patterns.

```
;; Was ok:
(parse-class "a b") := ["a" "b"]
(parse-class ["a" "b"]) := ["a" "b"]

;; Was failing - now fixed
(parse-class ["a b" "c"]) := ["a b" "c"] ; "a b" is not a valid dom class (contains a space)
```

For cross browser compat, some DOM attributes should always be set using
`setAttribute`, not by their corresponding property¹. E.g.

```javascript
node.setAttribute("colspan", 2); // no browser quirks
node.colSpan = 2;                // potential browser quirks
```

The Google Closure Library lists these attributes in `goog.dom.DIRECT_ATTRIBUTE_MAP_`².
We respect this requirement and implemented a similar logic to the one in
`goog.dom`, which given a `colspan` property will call
`setAttribute("colSpan")`. We later introduced a bug when we added SVG support
which requires the newer `setAttributeNS` method. `setAttributeNS`
differentiates between `colspan` vs `colSpan`, where the older `setAttribute`
doesn't. Thus, not setting the right attribute.

¹: https://github.com/google/closure-library/blob/c31b49fecdbea74df02049358460deda64bff244/closure/goog/dom/dom.js#L519
²:

```javascript
/**
 * Map of attributes that should be set using
 * element.setAttribute(key, val) instead of element[key] = val.  Used
 * by goog.dom.setProperties.
 *
 * @Private {!Object<string, string>}
 * @const
 */
goog.dom.DIRECT_ATTRIBUTE_MAP_ = {
  'cellpadding': 'cellPadding',
  'cellspacing': 'cellSpacing',
  'colspan': 'colSpan',
  'frameborder': 'frameBorder',
  'height': 'height',
  'maxlength': 'maxLength',
  'nonce': 'nonce',
  'role': 'role',
  'rowspan': 'rowSpan',
  'type': 'type',
  'usemap': 'useMap',
  'valign': 'vAlign',
  'width': 'width'
};
```

- allows keywords and symbols
- returns only unique values
- allows `nil` punning
- fails fast on other values

upgrade

Add tests that blow up CI in case of regression

…e.CLICK)

By the time they ran they were running in a different context where
`this` wasn't the Window object. As a fix we compile the call as
interop, where the precondition holds.

Server used to send PING frames to client, which is more correct.
We want Jetty and HTTPKit compat. Jetty supports PING/PONG, not HTTPKit.

Server is now passive, and keeps the connection alive for some times after the
last received message. Client periodically sends the message "HEARTBEAT". Server
detects "HEARTBEAT" and discard it.

Previously threw exception for empty string or string with leading or
trailing spaces.

Moved to :dev alias :extra-deps

Used to be printed with println. We want uncaugh errors (bubbling to the reactor
root) to go to the default logger.

…imeouts

We don’t want foreign code to break our history impl by messing with HTML5
browser history state.

Default action is to navigate with the router in scope. When integrating in an
existing app with an existing router (e.g. fulcro app), we want to redirect
navigation to the existing router.

Clients were not reconnecting on server heartbeat timeout. Users would see a
blank page after waking up their computer from sleep or after switching to an
old browser tab.

On timeout, the server now closes the connection with code 1013 "Try again
later". Client will try again.

…ccess

REPL usu. print ex-data, but not prod log systems.
Putting ex-data at the end of the log message ensures it will be visible.

This metadata is used by CIDER, clojure-lsp, and possibly other editors to
support proper indentation of macros.

…ns for libraries. todo lift all datomic demos/contrib into electric-fiddle

…ontent matches, delete dead code

This reverts commit 3240013.

We don't see a good use case right now

Record lexical scope binding meta (e.g. ^:tag 'js) in env, for better interop code gen.

This reverts commit a4e8062.

Was tagged `^js` but is not a plain js object.

The originating object's metadata is passed into the method call's fn

fixes core.match usage, which can include a
clojure.lang.IPersistentVector class tag in certain patterns. This class
object ended up in the debug metadata and cljs could not serialize it.

WARN: keeping arbitrary metadata is dangerous as there might be other
non-serializable values. We fixed this 1 edge case and left the rest
as-is for now, but a more general solution might be neede in the future.

The namespace object breaks cljs build.

fixes clojure.tools.logging

see https://clojurians.slack.com/archives/C8V0BQ0M6/p1690840683128579

…ore-async as :skip-doc