W3C home > Mailing lists > Public > public-html@w3.org > August 2007

The canvas element (detailed review)

From: Philip Taylor <philip@zaynar.demon.co.uk>
Date: Sat, 04 Aug 2007 03:12:17 +0100
Message-ID: <46B3E081.1070505@zaynar.demon.co.uk>
To: public-html@w3.org

I read through most of the canvas section some months ago outside the 
HTML WG, so now I'm just finishing up the bits I didn't get to before. 
(If anyone else is looking at this section, there are a number of 
unresolved issues I've already raised on the WHATWG list:

http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-June/011799.html
- shadows
http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-June/011814.html
- ImageData
http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-June/011883.html
- line styles
http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-June/011952.html
- patterns
http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-June/012103.html
- rectangles
http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-July/012117.html
- arcTo
http://lists.whatwg.org/htdig.cgi/whatwg-whatwg.org/2007-July/012126.html 
- arc

(My comments on the earlier parts have been satisfied by updates to the 
spec.))


fill:

"The fill() method must fill each subpath of the current path in turn" - 
that sounds incorrect - in current implementations it only paints one 
shape, which is made from all the subpaths at once. (If two overlapping 
but otherwise independent subpaths have opposite windings, they cancel 
out and result in no fill. If they have the same winding, that area just 
gets painted once.)

"the non-zero winding number rule" - that isn't defined, and it's not 
obvious what it means, so I think it should be defined here. 
http://www.cs.rit.edu/~icss571/filling/alt_parity.html and 
http://www.w3.org/TR/SVGMobile12/painting.html#FillRuleProperty sound 
like reasonable explanations.


stroke:

"The stroke() method must stroke each subpath of the current path in 
turn" - that sounds incorrect too - it only paints one shape, which is 
the union of all the subpaths' stroke shapes.


clip:

"The clip() method must create a new clipping path [...]" - doesn't say 
what should be done with the new clipping path that has been created.

"[...] a new clipping path [...]" - doesn't use the term "clip region", 
which is used elsewhere - the terminology should be consistent. "Clip 
region" sounds a better term since the interesting thing is the region 
that is clipped to, not the path that defines that region; and there 
isn't even an easy-to-compute path once you've started intersecting 
various regions. So it could say "The clip() method must set the current 
clip region to be the intersection of the current clip region and the 
area described by the current path (after applying the current 
transformation) using the non-zero winding number rule."

"When the context is created, the initial clipping path is the rectangle 
with the top left corner at (0,0) and the width and height of the 
coordinate space." - I don't think that's precise about the cases when 
the canvas's size is changed and the context is not being created but is 
"cleared back to [its] initial state and reinitialised with the newly 
specified coordinate space dimensions". It would be clearer to say "When 
the context is initialised, the clipping path is [...]", so it's obvious 
that that's relevant to reinitialisation too.

"When the context is created, the initial clipping path is the rectangle 
[...]" - should use "must".


isPointInPath:

Is a point on the edge of a path considered inside or out or undefined? 
(When doing a rect() path, FF3 says all points on the edge except the 
top-left corner are in. Opera 9.5 says all points on the bottom edge or 
right edge or bottom-right corner are in.)

Should the clip region be taken into account? FF3 and O9.5 ignore it, 
which seems logical since this is "is point in path" and not "is point 
in the area that would actually get rendered if you filled (but don't 
count shadows and alpha)". It could say "The isPointInPath(x, y) method 
must return true if the point [...] is inside the current path (after 
applying the current transformation), using the non-zero winding number 
rule; [...]."

The Transformations section says "The transformation matrix is applied 
to all drawing operations prior to their being rendered. It is also 
applied when creating the clip region" - should add "and when invoking 
the isPointInPath() method."

"canvas' coordinate space" - that is written "canvas coordinate space" 
elsewhere (which is better since I wouldn't be tempted to complain about 
"canvas'" vs "canvas's").


drawImage:

"The image argument must be an instance of an HTMLImageElement or 
HTMLCanvasElement. If the image is of the wrong type, the implementation 
must raise a TYPE_MISMATCH_ERR exception." - what if it is the correct 
type but null? (The same applies to createPattern). (Currently, FF 
throws NS_ERROR_NOT_AVAILABLE, Opera throws TYPE_MISMATCH_ERR, Safari 
crashes.)

What should happen if sw = 0 or sh = 0? (Firefox draws nothing, Opera 
(9.2) draws something (top left pixel?), Safari 3 throws INDEX_SIZE_ERR.)

What should happen if the image is animated? (Firefox and Safari draw 
the current frame. Opera draws nothing). Similar issue with 
createPattern. (In that case, Firefox 2 and Safari always draws the same 
frame as was first drawn; Firefox 3 randomly (each time you reload the 
page) either does the same as FF2 or else uses the current animation 
frame each time it draws something with the pattern; Opera returns 
undefined from createPattern.)

It might be nice to say "Implementations should use bilinear filtering 
when resizing the image", so that implementations agree with each other 
except when they have reasons not to (e.g. resource constraints which 
make bilinear filtering too expensive). (At least, I think current 
implementations are doing bilinear filtering - is that correct?)

It's not explicit that the transformation applies only to the 
destination rectangle, and not the source rectangle.

"If not specified, the dw and dh arguments default to [...]. If the sx, 
sy, sw, and sh arguments are omitted, they default to [...]" - it would 
be more convenient to test those cases if they said "must".

"If one of the sy, sw, sw, and sh arguments is outside the size of the 
image [...]" - should be "sx, sy, sw, and sh", though it may be better 
to say "If the source rectangle defined by the sx, sy, sw, and sh 
arguments is outside the image [...]".

"the specified region of the image specified by the source rectangle" -> 
"the region of the image specified by the source rectangle"

-- 
Philip Taylor
philip@zaynar.demon.co.uk
Received on Saturday, 4 August 2007 02:12:20 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 9 May 2012 00:16:03 GMT