Bug 9673 – Add support for window.atob() and window.btoa()

• Component: DOM
• Status: RESOLVED FIXED

[Alexey Proskuryakov]

These convert between base64 and binary strings, see the bug URL for documentation.

Supported by Mozilla and Konqueror, not supported by Opera 9, haven't tested WinIE.

[Alexey Proskuryakov]

Created attachment 9117 [details]
btoa test

[Alexey Proskuryakov]

Created attachment 11405 [details]
proposed fix

This implementation is based on the one from KDE, but has been tweaked for better Firefox compatibility.

One thing I couldn't understand in khtml (and thus didn't copy) was a special case for AToB/BToA in Window::getOwnPropertySlot:

    switch(entry->value) {
    case Frames:
// ...
    case Blur:
    case AToB:
    case BToA:
      getSlotFromEntry<WindowFunc, Window>(entry, this, slot);

<http://websvn.kde.org/trunk/KDE/kdelibs/khtml/ecma/kjs_window.cpp?rev=592458&view=markup>

WebCore has similar (but not identical) code for Blur etc. here, so maybe we actually need it, too.

[Sam Weinig]

Comment on attachment 11405 [details]
proposed fix

Setting r- due to a couple of issues I see.  

In Base64.cpp:

The * should be next to the type

+inline int strncasecmp(const char *first, const char *second, size_t maxLength)

Though, you should probably use the version of strncasecmp in StringExtras.h if you can.

Please use "unsigned" instead of "unsigned int" in a bunch of places.

Put spaces around infix operators in bunch of places as well.

Also, since this is based on the KDE impl, if there is a bug number or a patch that implemented this it would be nice to have a link to it or at least note who the author of the   patch was.

[Alexey Proskuryakov]

Created attachment 11425 [details]
proposed fix

> Though, you should probably use the version of strncasecmp in StringExtras.h if
> you can.

  Done (had to add the header to wtf, however - hope this is OK). Also made this change in DeprecatedString.cpp.

> Please use "unsigned" instead of "unsigned int" in a bunch of places.

  Do we have such a rule? I don't see one in the guidelines, and I don't like this style myself.

> Put spaces around infix operators in bunch of places as well.

  OK, but please note that I intentionally broke this rule in these places - I think adding spaces there has a negative impact on redability. But done.

> Also, since this is based on the KDE impl, if there is a bug number or a patch
> that implemented this it would be nice to have a link to it or at least note
> who the author of the   patch was.

  Base64 implementation comes from kdecore and is completely unrelated to khtml. Its authorship is already reflected in the copyright header.

  The kjs_window part of this implementation is trivial (but note the getOwnPropertySlot question I asked in comment 2). I haven't researched its history.

[Alexey Proskuryakov]

(In reply to comment #4)
> (but note the getOwnPropertySlot question I asked in comment 2). 

  OK, I think I see it now. The code is there to allow/deny access to functions from other windows (e.g. window.frames[0].close). Logically speaking, atob() and btoa() are absolutely safe, so there's no reason to disallow them - but Firefox does that, so the patch stands as it is.

[Geoffrey Garen]

> > Please use "unsigned" instead of "unsigned int" in a bunch of places.
> 
>   Do we have such a rule? I don't see one in the guidelines, and I don't like
> this style myself.

I'm not sure if it's officially in the guidelines, but Darin harps on it from time to time. I guess being concise is his argument.

[Alexey Proskuryakov]

Created attachment 11770 [details]
now with unsigned instead of unsigned int

OK. Would be nice to put this into official guidelines then!

[Darin Adler]

Comment on attachment 11770 [details]
now with unsigned instead of unsigned int

Looks good. r=me

[Alexey Proskuryakov]

Committed revision 18170.