Add feComposite support for operator attribute's lighter value #9074

hamishwillee · 2021-02-12T05:54:20Z

FF86 introduces support for the "lighter" filter attribute - see bug 1518099.

The current BCD on feComposite is a bit of a mess - for example it says that this is already supported in FF and is not supported in Chrome, neither of which is true.

Bug info for main platforms here:

Chrome - release 45, which I got from here. I have rolled out that support to equivalent versions for Opera, Samsung etc.
Firefox - FF86
Webkit 141376 - fixed in v528+

Following on from those bugs it is pretty clear that feComposite must be supported in Blink and Webkit derived browsers. Have marked as true. For Blink have marked as version version where Blink was supported.

Docs bug is mdn/content#1727

This may also affect SVGFECompositeElement but still checking.

hamishwillee · 2021-02-12T06:01:47Z

svg/elements/feComposite.json

@@ -24,16 +24,16 @@
              "version_added": null
            },
            "opera": {
-              "version_added": false
+              "version_added": true


Note, these must all be true/supported because you can't have support for "lighter" if the parent item is not supported (as per bugs). I am not sure what versions.

If you believe chrome then this was in from release 1 (webkit and from then in Blink). Should I leave these alone or assume that support came in earliest matching version of Blink/Webkit for the other browsers? That would mean, for example, that Opera would change to version 15 and I'd assume there was no support in Presto. Thoughts?

svg/elements/feComposite.json

hamishwillee · 2021-02-16T01:03:29Z

api/SVGFECompositeElement.json

@@ -51,40 +51,52 @@
        "__compat": {
          "support": {
            "chrome": {
-              "version_added": "5"
+              "version_added": "5",
+              "notes": "BackgroundImage/BackgroundAlpha are not supported (see <a href='https://crbug.com/137230'>Chromium bug 137230</a>)."


FYI added all these "BackgroundImage/BackgroundAlpha notes to mirror feComposite (which this API does). Only thing that supports it is Edge before v79

A few thoughts there:

If I understand correctly, this note means that neither BackgroundImage nor BackgroundAlpha are accepted as values to the in1 or in2 attributes?

Ordinarily, I'd say that when we repeat the same note again and again, it's often a sign that the information needs to be expressed as a subfeature or something else odd is going on. Except, in this case…

~~Apparently no browser supports these values, so there's no actual compat story to worry about, right?~~

~~Unless there's a reason for web developers to expect these values to work, I'm inclined to suggest dropping all of these identical notes.~~

Oh, wait, I get it now: Edge before Edge 79 is the oddity; no current browser supports it, but there was somewhat recent support in Edge. Nobody else has implemented this. Reading the linked bugs, it looks like nobody else will implement it.

Here's my new take:

Drop all the notes on the SVGFECompositeElement data about accepted attributes. in1 etc. are reflected attributes, so we don't cover supported values in the interface. I get into more detail about this on the lighterForError feature, since it's a bigger issue there.

On the feComposite data, reduce this to one and only one note, for Edge (see suggestion below).

Essentially yes. You can specify those of course (they are "valid" options) but the only thing they they work on is pre-Blink-Edge. So "no current browser".

My read of the defect reports is because this is very hard to do, not because it isn't a good idea. Even the example in the SVG spec uses these. I have just removed that example from the docs. Also yes, that some browsers won't ever fix this.

But my concern is that without the note everyone will think all the spec values for in1 and in2 will work. I think you're saying "that doesn't matter". I will take your direction and update appropriately :-)

To clear up a point of confusion: reflected attributes set and get the source element's attributes. For example, given page source <feComposite in1="somevalue">, SVGFECompositeElement.in1 returns "somevalue" and SVGFECompositeElement.in1 = "NEW" sets the element to <feComposite in1="NEW">. In terms of behavior, reflected attributes provide access to underlying element but don't really add anything more.

Consequently, to avoid duplication, we document support for attribute values on the element itself, not the interface to that element. We figure that people don't really learn about, for example, the attributes of HTML elements by reading the HTML*Element interface docs.

my concern is that without the note everyone will think all the spec values for in1 and in2 will work

I didn't realize the spec used these (unsupported) values as examples and I appreciate that you're trying to guard against misunderstanding here. In that case, I think the right way to handle this is to have subfeatures for in1 and in2 (on the svg.elements.feComposite.{in1,in2} features) that represent each of the values for it and provide complete compat data for each, instead of notes. I think that gives readers a clearer idea of what is and is not supported.

There's one additional bit of nuance here, however: because spec authors write ahead of implementation, there are often features which are in specs and never implemented. BCD can be a source of additional confusion, as folks get the idea that a feature in BCD means that it's a live concern. This is why we avoid all false features, where possible.

In this case, the BackgroundImage and BackgroundAlpha values were once supported by at least one browser, so that deserves an entry.

I was trying to avoid repetitious notes. Repetitious notes are hard to follow. When every entry has a note and the text is very similar between them, then it's hard to know what's actually noteworthy. That's why I was leaning toward writing notes for the exceptional, not usual, case.

Two additional thoughts here:

I missed that IE supported BackgroundImage and BackgroundAlpha as well. I think my general point still stands, however, that these are interesting because they're different from the others.

I said that we should "have subfeatures for in1 and in2 (on the svg.elements.feComposite.{in1,in2} features) that represent each of the values". I am not asking you to add features for every value yourself as part of this PR (that's optional). I just meant that svg.elements.feComposite.{in1,in2}.BackgroundImage and svg.elements.feComposite.{in1,in2}.BackgroundAlpha would be germane here.

@ddbeck You didn't miss anything about IE. I'm an idiot - those notes were supposed to be on Edge. Fixed.

But my feeling is that this is probably worth merging now because it is "not wrong" and adds a lot of corrected information about supported versions.

There are ongoing discussions like #9074 (comment) and this one on possibly adding subfeatures, but I'd happily do those as a post process. Up to you.

hamishwillee · 2021-02-16T01:06:25Z

api/SVGFECompositeElement.json

@@ -329,6 +351,54 @@
          }
        }
      },
+      "lighterForError": {


Robert (domain expert) confirmed lighterForError (what a weird key name for operator supporting the lighter value!) should be added to SVGFECompositeElement too. I have matched the feComposite versions exactly except for Safari.

For Safari the operator was not added until version 6, so I pushed the implementation of operator value lighter to be that same version. That is not "rock solid" (i.e. no bug confirmation) but very likely.

Sorry, but we should drop this feature entirely. For reflected attributes, we typically just track whether the attribute exists as part of the API. This is us mostly following the convention of specifications for these things: the IDL usually specifies the reflection, while the specification for the element defines valid values.

(Plus—and this is definitely not your fault, since you're following the structure from the element data—the structure is wrong. This ought to be a subfeature of operator, not a peer to operator.)

@ddbeck So to summarize, I just remove this addition altogether? - we simply don't track things supported at this level in the BCD. Sounds nuts, but there you go.

Yes, we should only have data on the element, not the interface. I got into this more in #9074 (comment). But let me know if you have more questions—it's not an obvious distinction.

@ddbeck What I think you're saying is that we don't document the bits of an SVG element that reflects a javascript API - we'd just document the javascript behaviour. Right?

So I understand why you wanted me to delete the lighterForError from the reflected API svg/elements/feComposite.json , but THIS is api/SVGFECompositeElement - the API from which we reflect. Which makes me think that it should have stayed.

Or did you mean that we document on the element but not the javascript API? Either way, doesn't one of them need comments about the features? The comments here said delete this from both /api and /svg

What am I missing?

Or did you mean that we document on the element but not the javascript API? Either way, doesn't one of them need comments about the features? The comments here said delete this from both /api and /svg

I meant on the element, not the API. Sorry, I didn't mean to suggest deleting all of them everywhere. I think the various threads on this PR made it hard to keep track of which thing was which. I'll post a standalone review that summarizes things that need to change to get this merged.

hamishwillee · 2021-02-16T01:15:31Z

api/SVGFECompositeElement.json

+            "firefox_android": {
+              "version_added": "86"
+            },
+            "ie": {


Note, IE doesn't work support lighter for feComposite, so set false in that and here too (verified by testing).

svg/elements/feComposite.json

hamishwillee · 2021-02-16T01:23:47Z

@evilpie @ddbeck This should now be ready for review. I've added notes throughout on what I've done. General things to note to make review easier are that according to Robert, the Moz domain expert:

if feComposite is supported, then in, in2 and all the k values are supported by everything. So these are all moved from null to at least true (if version not known).
feComposite and SVGFECompositeElement mirror each other in terms of functionality. That's why I updated that to have lighter and BackgroundImage. Versions might not completely match up though.
BackgroundImage/BackgroundAlpha only supported by Edge prior to Blink support.

…nto feComposite

ddbeck

Wow, this is a complicated one! Thanks for digging into this, @hamishwillee. I've got a bunch of suggestions, but it took me a while to understand the whole story here, so it might be a bit confused in places. If there's any questions about my comments, please say something. Thank you!

ddbeck · 2021-02-16T12:27:51Z

api/SVGFECompositeElement.json

@@ -51,40 +51,52 @@
        "__compat": {
          "support": {
            "chrome": {
-              "version_added": "5"
+              "version_added": "5",
+              "notes": "BackgroundImage/BackgroundAlpha are not supported (see <a href='https://crbug.com/137230'>Chromium bug 137230</a>)."


A few thoughts there:

If I understand correctly, this note means that neither BackgroundImage nor BackgroundAlpha are accepted as values to the in1 or in2 attributes?

Ordinarily, I'd say that when we repeat the same note again and again, it's often a sign that the information needs to be expressed as a subfeature or something else odd is going on. Except, in this case…

~~Apparently no browser supports these values, so there's no actual compat story to worry about, right?~~

~~Unless there's a reason for web developers to expect these values to work, I'm inclined to suggest dropping all of these identical notes.~~

svg/elements/feComposite.json

ddbeck · 2021-02-16T12:44:23Z

api/SVGFECompositeElement.json

@@ -51,40 +51,52 @@
        "__compat": {
          "support": {
            "chrome": {
-              "version_added": "5"
+              "version_added": "5",
+              "notes": "BackgroundImage/BackgroundAlpha are not supported (see <a href='https://crbug.com/137230'>Chromium bug 137230</a>)."


Oh, wait, I get it now: Edge before Edge 79 is the oddity; no current browser supports it, but there was somewhat recent support in Edge. Nobody else has implemented this. Reading the linked bugs, it looks like nobody else will implement it.

Here's my new take:

Drop all the notes on the SVGFECompositeElement data about accepted attributes. in1 etc. are reflected attributes, so we don't cover supported values in the interface. I get into more detail about this on the lighterForError feature, since it's a bigger issue there.

On the feComposite data, reduce this to one and only one note, for Edge (see suggestion below).

svg/elements/feComposite.json

ddbeck · 2021-02-16T13:00:13Z

api/SVGFECompositeElement.json

@@ -329,6 +351,54 @@
          }
        }
      },
+      "lighterForError": {


Sorry, but we should drop this feature entirely. For reflected attributes, we typically just track whether the attribute exists as part of the API. This is us mostly following the convention of specifications for these things: the IDL usually specifies the reflection, while the specification for the element defines valid values.

(Plus—and this is definitely not your fault, since you're following the structure from the element data—the structure is wrong. This ought to be a subfeature of operator, not a peer to operator.)

svg/elements/feComposite.json

hamishwillee · 2021-02-19T06:49:06Z

@ddbeck Thanks for the review. I think I have now done the fixes. Hope I got it right.

hamishwillee · 2021-02-19T06:49:33Z

PS Partially the response confused me because I don't know what a reflective property is.

ddbeck

Here's a summary of what needs to change before we merge this:

API features

Remove the notes (new in this diff) from the api features
Rename in1 to in (to match the name of the element's attributes and the spec—this is new and I only just noticed it)

SVG features

Reinstate lighterForError (SVG only, not API)
Move lighterForError one level deeper, as a child to operator
Rename lighterForError to lighter and rewrite the description to <code>lighter</code> value

ddbeck · 2021-02-23T08:32:57Z

api/SVGFECompositeElement.json

@@ -57,7 +57,8 @@
              "version_added": "18"
            },
            "edge": {
-              "version_added": "12"
+              "version_added": "12",
+              "notes": "Before Edge 79, <code>BackgroundImage</code> and <code>BackgroundAlpha</code> were supported."


Remove this note (the one on svg.elements.feComposite.in covers it)

svg/elements/feComposite.json

api/SVGFECompositeElement.json

hamishwillee · 2021-02-26T03:11:04Z

Thanks for the clear direction @ddbeck . I'm pretty sure this captures it.

chrisdavidmills · 2021-03-02T10:34:28Z

@ddbeck this one is mentioned in the 86 rel notes: https://developer.mozilla.org/en-US/docs/Mozilla/Firefox/Releases/86#svg

Can you have a quick look at @hamishwillee 's latest fixes and get it merged if you are happy? Thanks!

ddbeck

Thank you! 🎉

Original source: mdn#9074 This demo was used to test: https://yari-demos.prod.mdn.mozit.cloud/en-US/docs/Web/SVG/Element/feComposite/_sample_.example.html The following support was confirmed with BrowserStack: - Chrome 41 (not in 40) - Edge 79 (not in 18) - Firefox 86 (not in 85) - Safari 10.1 (not in 9.1) The precise Safari version was determined by source: https://trac.webkit.org/changeset/195745/webkit https://trac.webkit.org/browser/webkit/trunk/Source/WebCore/Configurations/Version.xcconfig?rev=195745 WebKit trunk version 602.1.18 maps to Safari 10 (using WebKit 602.1.50). Part of mdn#7844.

foolip · 2022-05-13T11:42:51Z

Sorry to revive an old PR, but just a note that I found the Chrome, Edge and Safari versions here to be incorrect and fixed them in #16278. For Chromium and WebKit, it appears that commits were mapped to versions in an incorrect way. The Chrome mistake is probably due to #7844, but I don't know what happened for WebKit.

feComposite support for operator lighter property

e5895f5

github-actions bot added the data:svg Compat data for SVG features. https://developer.mozilla.org/docs/Web/SVG label Feb 12, 2021

hamishwillee commented Feb 12, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee mentioned this pull request Feb 12, 2021

[SVG] Implement <feComposite operator="lighter"> mdn/content#1727

Closed

hamishwillee added 2 commits February 15, 2021 10:50

Fix up bug link for images

b0b2068

Improve supported info

5b49a95

hamishwillee commented Feb 15, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee commented Feb 15, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee commented Feb 15, 2021

View reviewed changes

svg/elements/feComposite.json Show resolved Hide resolved

hamishwillee commented Feb 15, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee commented Feb 15, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee mentioned this pull request Feb 15, 2021

SVG feComposite improvements mdn/content#2353

Merged

Further updates to feComposite and SVGFECompositeElement

2378817

github-actions bot added the data:api Compat data for Web APIs. https://developer.mozilla.org/docs/Web/API label Feb 16, 2021

hamishwillee commented Feb 16, 2021

View reviewed changes

More minor fixes

bb30e24

hamishwillee commented Feb 16, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee commented Feb 16, 2021

View reviewed changes

svg/elements/feComposite.json Outdated Show resolved Hide resolved

hamishwillee commented Feb 16, 2021

View reviewed changes

svg/elements/feComposite.json Show resolved Hide resolved

Merge branch 'master' of https://github.com/mdn/browser-compat-data i…

5199c46

…nto feComposite

ddbeck requested changes Feb 16, 2021

View reviewed changes

hamishwillee added 3 commits February 19, 2021 17:32

Fix up versions and remove ligherforerror

d2946a6

Remove lighterforerror from SVGFECompositeElement (reflective)

ef70989

BackgroundImage note just for IE

81b2f5a

Edge not IE for note

722ea03

ddbeck requested changes Feb 23, 2021

View reviewed changes

hamishwillee added 2 commits February 26, 2021 13:42

remove notes from SVGFECompositeElement (in, in2)

dcdccb2

Add back lighter:

ddf02e4

Fix up descript as requested

644bddc

ddbeck added the needs-release-note 📰 label Mar 1, 2021

ddbeck approved these changes Mar 3, 2021

View reviewed changes

ddbeck merged commit 40af4a0 into mdn:master Mar 3, 2021

ddbeck changed the title ~~feComposite support for operator lighter property~~ Add feComposite support for operator attribute's lighter value Mar 3, 2021

ddbeck added a commit to ddbeck/browser-compat-data that referenced this pull request Mar 3, 2021

Start release notes for v3.1.3 with feature rename from mdn#9074

1000da6

ddbeck added a commit to ddbeck/browser-compat-data that referenced this pull request Mar 3, 2021

Start release notes for v3.1.3 with feature rename from mdn#9074

9b25335

ddbeck added a commit to ddbeck/browser-compat-data that referenced this pull request Mar 4, 2021

Start release notes for v3.1.3 with feature rename from mdn#9074

d4e14f8

ddbeck added a commit to ddbeck/browser-compat-data that referenced this pull request Mar 4, 2021

Start release notes for v3.1.3 with feature rename from mdn#9074

b73212e

saschanaz mentioned this pull request Mar 4, 2021

api.SVGFECompositeElement.in - Name doesn't match with Web IDL #9337

Closed

dependabot bot mentioned this pull request Mar 7, 2021

Bump @mdn/browser-compat-data from 3.1.2 to 3.1.3 mrhenry/core-web#420

Merged

hamishwillee deleted the feComposite branch March 8, 2021 05:47

queengooborg removed the needs-release-note 📰 label Apr 24, 2022

foolip mentioned this pull request May 13, 2022

Update browser versions for <feComposite operator="lighter"> #16278

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add feComposite support for operator attribute's lighter value #9074

Add feComposite support for operator attribute's lighter value #9074

hamishwillee commented Feb 12, 2021 •

edited

Loading

hamishwillee Feb 12, 2021

hamishwillee Feb 16, 2021 •

edited

Loading

ddbeck Feb 16, 2021

ddbeck Feb 16, 2021

hamishwillee Feb 19, 2021

ddbeck Feb 19, 2021

ddbeck Feb 19, 2021

hamishwillee Feb 23, 2021

hamishwillee Feb 16, 2021 •

edited

Loading

ddbeck Feb 16, 2021

hamishwillee Feb 19, 2021 •

edited

Loading

ddbeck Feb 19, 2021

hamishwillee Feb 23, 2021 •

edited

Loading

ddbeck Feb 23, 2021

hamishwillee Feb 16, 2021

hamishwillee commented Feb 16, 2021

ddbeck left a comment

ddbeck Feb 16, 2021

ddbeck Feb 16, 2021

ddbeck Feb 16, 2021

hamishwillee commented Feb 19, 2021

hamishwillee commented Feb 19, 2021

ddbeck left a comment •

edited

Loading

ddbeck Feb 23, 2021

hamishwillee commented Feb 26, 2021

chrisdavidmills commented Mar 2, 2021

ddbeck left a comment

foolip commented May 13, 2022

Add feComposite support for operator attribute's lighter value #9074

Add feComposite support for operator attribute's lighter value #9074

Conversation

hamishwillee commented Feb 12, 2021 • edited Loading

Choose a reason for hiding this comment

hamishwillee Feb 16, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hamishwillee Feb 16, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hamishwillee Feb 19, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hamishwillee Feb 23, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hamishwillee commented Feb 16, 2021

ddbeck left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hamishwillee commented Feb 19, 2021

hamishwillee commented Feb 19, 2021

ddbeck left a comment • edited Loading

Choose a reason for hiding this comment

API features

SVG features

Choose a reason for hiding this comment

hamishwillee commented Feb 26, 2021

chrisdavidmills commented Mar 2, 2021

ddbeck left a comment

Choose a reason for hiding this comment

foolip commented May 13, 2022

hamishwillee commented Feb 12, 2021 •

edited

Loading

hamishwillee Feb 16, 2021 •

edited

Loading

hamishwillee Feb 16, 2021 •

edited

Loading

hamishwillee Feb 19, 2021 •

edited

Loading

hamishwillee Feb 23, 2021 •

edited

Loading

ddbeck left a comment •

edited

Loading