Brian Hung

> I've [proposed to use special types](https://github.com/mobxjs/mobx-state-tree/issues/1282#issuecomment-515026561) (`types.arrayReference`/`types.mapReference`) - moving to those is a _breaking change_, but will allow us to remove `identifiersCache` completely (which also gives us a chance...

Confused how to use the undoManager since it runs into this exact issue; can't addMiddleware to root store if it doesnt exist when child is created. https://github.com/mobxjs/mobx-state-tree/blob/bac4349daa2cc2cd3e2e3516ba62f5d4ea866c27/packages/mst-middlewares/src/undo-manager.ts#L135

Is it possible to have a minor release with the lodash module fix? Running into deployment issues with out of memory from lodash being entirely bundled. @plouc

This seems like a possible solution https://github.com/GoogleChromeLabs/comlink/issues/450#issuecomment-827503286 > I’d recommend just transferring to MessageChannels, one for each direction of communication. Does anyone know if we can use BroadcastChannels to avoid...

#294 has an example of using `Comlink.createEndpoint` to indirectly use a MessageChannel. Think this issue can be closed now.

Also think a development only error message or warning would be great, since it requires a bit of digging to find that you cant pass non-serializable state into navigate.

@sequba Ran into a similar issue with batching cells with the same address when evaluation is suspended: https://codesandbox.io/s/young-butterfly-zdxrvc?file=/src/hyperformulaConfig.js ```ts import HyperFormula from "hyperformula"; console.log( `%c Using HyperFormula ${HyperFormula.version}`, "color: blue;...

Reposting from discussions: https://github.com/handsontable/hyperformula/discussions/1155 #845 #135 https://github.com/handsontable/handsontable/issues/8114 It would be useful to have a `formulasUpdated` listener, as mentioned in #135. I've been trying to integrate hyperformula with an external data...

> a new model that was selfless oh in hindsight I definitely should’ve just created a separate model 😅 > my initial reading correct to say this is a fully...

@k-g-a Would it be possible to post the proof of concept implementation? Would love to read that!