Conditions

• confirmed this problem with examples/locations.html.
  • This is my code. https://gist.github.com/tjmtmmnk/87742bc8106a89ab3d018b28f341eb86
• checked using the epub "Vertical Writing" provided on this site. http://idpf.github.io/epub3-samples/30/samples.html
• checked by pixel5 on the developer tools

Problem

Offset is shifted when calling display(cfi).

Possible Causes

• When calling display(cfi), DefaultViewManger.display -> DefaultViewManager.moveTo is executed and offset is adjusted.
• Looking at DefaultViewerManager.next, it executes scrollBy for this.layout.delta in the x direction when ltr and this.layout.height in the y direction when rtl and vertical.

  • epub.js/src/managers/default/index.js

    Line 466 in 0963efe

    this.scrollBy(this.layout.delta, 0, true);

  • epub.js/src/managers/default/index.js

    Line 499 in 0963efe

    this.scrollBy(0, this.layout.height, true);

• On the other hand, for moveTo, if rtl and vertical, it executes scrollTo for distY = Math.floor(offset.top / this. layout.delta) * this.layout.delta;

  • epub.js/src/managers/default/index.js

    Line 353 in 0963efe

    distY = Math.floor(offset.top / this.layout.delta) * this.layout.delta;

  • Therefore, it is not a constant multiple of this.layout.height and is considered to be shifted.

countermeasure

I believe we need to change this.layout.delta in distY = Math.floor(offset.top / this.layout.delta) * this.layout.delta; to this.layout.height.

- distY = Math.floor(offset.top / this.layout.delta) * this.layout.delta;
- if (distY + this.layout.delta > this.container.scrollHeight) {
-  distY = this.container.scrollHeight - this.layout.delta;
}

+ distY = Math.floor(offset.top / this.layout.height) * this.layout.height;
+ if (distY + this.layout.height > this.container.scrollHeight) {
+   distY = this.container.scrollHeight - this.layout.height;
}

Results of measures

before	after

• Confirmed that the offset shift was gone.
• I also confirmed that for horizontal epub, offset.top < this.layout.delta always holds and distY=0, with no effect.

If this change is considered reasonable I would like to submit a PR. Thank you!