Closed Bug 393910 Opened 16 years ago Closed 11 years ago

Serialize "0" as "0px" for lengths instead of "0pt"


(Core :: CSS Parsing and Computation, enhancement)

Not set





(Reporter: jruderman, Assigned: ayg)




(2 files, 2 obsolete files)

Attached file testcase
It would be nice if "width: 0" would stay as "0" instead of becoming "0pt" when serialized.  I rarely use pt units so it's strange to see them appear unnecessarily.
IE7/Windows gives 0px
Opera 9.23/Linux gives 0
Konqueror gives 0px

I agree that 0 is better style; not sure if I want to make the switch at this point in the cycle.  Nobody else gives pt, though.
We could make a simple fix in nsCSSDeclaration::AppendCSSValueToString , but it would be nice to round-trip the unit as specified.  We currently store 0s as various default units in CSSParserImpl::TranslateDimension; the way to round-trip properly would involve adding a new unit for them and making sure that unit was handled appropriately.
I think this bug should really be about round-tripping what the author specified rather than converting other values to 0, though.

(The reason for points as the default in TranslateDimension may have been that points were cheaper to compute back when we used twips; that's no longer the case following bug 177805.)
OS: Mac OS X → All
Hardware: x86 → All

data:text/html,<!doctype html>
<script> = "0";
document.documentElement.textContent =

IE10 Developer Preview: 0px
mozilla-central:        0pt
Chrome 18 dev:          0px
Opera Next 12.00 alpha: 0px


"A length of zero is represented by the literal string "0px"."

Seems like "0px" is the winner here.  "0" makes more sense, but nobody does that right now, so it makes more sense to go with what everyone else does.
Attached patch Patch v1 (obsolete) — Splinter Review
This serializes it as "0px", like the spec and other browsers.
Assignee: nobody → ayg
Attachment #596762 - Flags: review?(dbaron)
Whiteboard: [autoland]
Whiteboard: [autoland] → [autoland-in-queue]
Autoland Patchset:
	Patches: 596762
	Branch: mozilla-central => try
Try run started, revision 56185fcd57d1. To cancel or monitor the job, see:
Comment on attachment 596762 [details] [diff] [review]
Patch v1

>+// Bug 393910
>+p.setAttribute("style", "margin-left: 0");
>+is("margin-left"), "0px",
>+  "0 serializes to 0px");
> for (var test in tests) {

Could you put your insertion after the main for loop instead of before?

r=dbaron with that
Attachment #596762 - Flags: review?(dbaron) → review+
Attached patch Patch v2 (obsolete) — Splinter Review
Attachment #596762 - Attachment is obsolete: true
Attachment #596770 - Flags: review?(dbaron)
Attachment #596770 - Flags: review?(dbaron) → review+
Try run for 56185fcd57d1 is complete.
Detailed breakdown of the results available here:
Results (out of 210 total builds):
    exception: 3
    success: 164
    warnings: 29
    failure: 14
Builds (or logs if builds failed) available at:
Whiteboard: [autoland-in-queue]
There was a test I missed in content/base/, because I only tested layout/style/ locally.
Attachment #596770 - Attachment is obsolete: true
Attachment #597038 - Flags: review+
Keywords: testcasecheckin-needed
Summary: Output unitless zero for style.width ("0" instead of "0pt") → Serialize "0" as "0px" for lengths instead of "0pt"
Flags: in-testsuite+
Keywords: checkin-needed
Target Milestone: --- → mozilla13
Closed: 11 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.