Give StaticRange a constructor

[foolip]

It's now immutable, but could still have a constructor:

dictionary StaticRangeInit {
  required Node startContainer;
  required unsigned long startOffset;
  required Node endContainer;
  required unsigned long endOffset;
};

[Constructor(StaticRangeInit initDict)]
interface StaticRange { ... };

If there's no use for invalid ranges, then it could throw if http://garykac.github.io/staticrange/index.html#to-is-valid does not hold.

[chong-z]

Do we have a resolution on what should happen if the range is invalid? IIUC the previous version of spec wants to do some normalization, e.g. It will tweak start and end to make the range valid.

@garykac Can you confirm the expected behavior here?

If there is no clear resolution or requires further discussion would it be better to ship StaticRange without a constructor or with only an empty constructor first?

[garykac]

@foolip What's the rule-of-thumb for adding constructors?

AIUI, DataTransfer originally didn't have one because there was no use case for it. That seems similar to the situation with StaticRange: We don't need one now, but we could add one later when there is a need.

Or should we default to adding a constructor even when there is no API need for one? I can't think of a use for it.

@choniong If we decide to add a constructor, then we would need to specify either (1) throw, (2) normalize like Ranges do. My pref is (2), but I don't think we need a constuctor.

[foolip]

The rule-of-thumb that I have inherited from @domenic and @slightlyoff is that things should have a constructor by default unless the objects are somehow "exotic" or cannot be created synchronously. ImageBitmap is an example where there are good reasons I think, but I'm not in touch with the details.

Here, I guess that the constructor should simply copy over the values from the init dictionary. Any StaticRange that the UA creates can be made invalid by modifying the DOM enough, so I don't think trying to prevent scripts from creating initially-invalid ranges would help.

[domenic]

Right. In general we are trying to minimize the amount of magic on the platform. Objects that the UA is magically able to construct somehow, despite the class in question not having a constructor, are serious magic. In some cases (e.g. ImageBitmap) that magic is necessary, but definitely not here.

[foolip]

@whsieh @rniwa, any concerns about adding a constructor in WebKit?

[rniwa]

Please just throw when the range is invalid then I don't see any reason adding a constructor is problematic.

[foolip]

I suppose throwing might catch some mistakes, so that seems fine to me, even if the range can later become invalid. I defer to @choniong and @garykac to say what they'd prefer.

[garykac]

I don't have strong feelings about this, so throwing is fine with me.

…

On Sun, Apr 16, 2017 at 9:25 PM, Philip Jägenstedt ***@***.*** > wrote: I suppose throwing might catch some mistakes, so that seems fine to me, even if the range can later become invalid. I defer to @choniong <https://github.com/choniong> and @garykac <https://github.com/garykac> to say what they'd prefer. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABaVSv8Zj7-QgTu6uGgKBvxUIqtB05RKks5rwxP5gaJpZM4M4avM> .

[garykac]

This issue was moved to w3c/staticrange#3