Improve the logic of isNode().

[tamuratak]

Improve the logic of isNode(). Close #320.

The trick of '[object process]' seems to be widely used.

• https://github.com/search?q=%22%5Bobject+process%5D%22&type=code
• https://github.com/mozilla/pdf.js/blob/7ea0e40e588864cd938d1836ec61f1928d3877d3/src/shared/util.js#L24

Now, since the process.versions is empty on jspm-core, you could close the original issue without merging this PR.

• https://github.com/jspm/jspm-core/blob/7af7d7413f472305d08d0d78ec3d1f15588be50a/nodelibs/browser/process.js#L78

[josdejong]

Thanks, this approach makes sense. How about removing the logic that checks .node.versions? I think that is redundant now. It can become just nodeProcess + "" === "[object process]", right ? Or String(nodeProcess) === "[object process]" to make it a bit more explicit.

> Now, since the `process.versions` is empty on `jspm-core`, you could close the original issue without merging this PR.

Do you mean that #320 _is_ already fixed without this PR?

[tamuratak]

How about removing the logic that checks .node.versions?

It would break workpool on the Electron renderer process, which has the built-in process although we should consider it as a browser platform.

• https://www.electronjs.org/docs/latest/api/process

Do you mean that #320 is already fixed without this PR?

Yes. This PR would be good for another polyfill library that will appear in the future.

[josdejong]

Ah, of course. In the isNodeJs function of pdf.js that you linked to there are also a couple of extra conditions to distinguish from Electron and NW I see. Ok let's keep the logic that we already had as is 👍 .

[josdejong]

Published now in v9.0.4, thanks again. I wish you nice holidays!