Workaround for #179 - use datalink for Chrome iOS

[ph4r05]

• Blob links / url objects are not supported in Chrome iOS for now

closes #179

[jimmywarting]

Not long ago i have switched the is_safari to test a webkit feature instead of doing browser sniffing on user agent

Safari's /constructor/i.test(view.HTMLElement) is uniq for Safari for Safari's webkit

And since Apple don't allow 3rd party browser engine chrome is basically stuck with running uiwebview or wkwebview
which is the same as running Safari basically.

So the conclusion here is: We can as well use is_safari instead of is_chrome_ios
Since chrome for ios also reports true for /constructor/i.test(view.HTMLElement)

[jimmywarting]

What those this really solve? see my comment. Chrome on ios should fall in to the same category as is_safari since it's the same rendering engine.

[ph4r05]

With this condition datalink is used in case of Chrome for iOS.
I can test your suggestion, to use only is_safari and I will let you know how it works on Safari and Chrome for iOS.

Whats your opinion on:

• opening a new page vs using previous with view.location.href. EDIT: Opening a new page with view.open(datalink, "_blank"); does not work in Safari, does work in CriOS.
• using blob.type instead of "data:attachment/file" - which does not work in Chrome iOS?

[ph4r05]

And I forgot, difference from is_safari is that for CriOS force == false thus datalink was not used

[ph4r05]

Also a big difference for Safari & CriOS - blob: object links are working in Safari, not in CriOS. (Also the reason why force == false for CriOS. It uses object links, which do not work).

[jimmywarting]

My opinion on opening a new page vs current page is: open a new page if possible otherwise use current page.

There is some scenarios where opening a new page won't work.
That is when a you try to open a new page that is made "to long ago" after a user interaction, since window.open() only allow to open on user interaction like onClick - default behavior is to block annoying popups otherwise
I have found out that you can run window.open() if you do it somewhere between 0-1000 ms after a user interaction occurred

So my preference would be

popup = window.open(x)
if(!popup) window.location.href = x

[jimmywarting]

Also a big difference for Safari & CriOS - blob: object links are working in Safari, not in CriOS. (Also the reason why force == false for CriOS. It uses object links, which do not work).

So what your commit really comes down to is this?

force = type === force_saveable_type || /CriOS\/[\d]+/.test(navigator.userAgent)

[ph4r05]

force = type === force_saveable_type || /CriOS\/[\d]+/.test(navigator.userAgent)

Almost, but not exactly I think. If force == true the datalink is used, thats great.

1. difference is to open it in a new tab in CriOS - approach you suggested recently works fine.
2. Datalink with "data:attachment/file" does not work in CriOS. But datalink with "data:" + blob.type or "data:application/octet-stream" worked for me in CriOS.

[jimmywarting]

Just a observation, what happens if the blob don't have a type?
i.e. new Blob(["hello world"])

[ph4r05]

Nice one! It is empty.
https://jsfiddle.net/jv1fhkst/

I've just tested the empty datatype also works in CriOS.

Or we can then use:

blob.type || "application/octet-stream"

Both display plaintext "hello world" in a new tab.

[jimmywarting]

Feels like we could do something smarter... we are basically reconstructing the same result we get from the reader if it's CriOS? So something like this would be better?

var url = is_chrome_ios ? reader.result : reader.result.replace(/mimetype/, 'attachment/file')
var popup = view.open(url, '_blank')
if(!popup) view.location.href = url

(regex is wrong, but a quick google search may find something)

Something along that line... Probably should use view instead of window also

[ph4r05]

• Regarding view instead of window - sure! it was a bug. my fault.
• Regarding url - will let you know after I test it

[ph4r05]

It works quite nice! thanks for suggestion

[ph4r05]

Once we have a solution, I can squash it to a single commit so it does not mess the commit history if you want.

[eligrey]

lgtm

[ph4r05]

squashed, minified version updated