From a86d830d0be4da5d6d161ad9e053660ccc37fd2f Mon Sep 17 00:00:00 2001
From: Adrian Cochrane <adrian@openwork.nz>
Date: Mon, 17 Jun 2019 13:29:53 +1200
Subject: [PATCH] Document plans for supporting pseudo-classes/elements.

---
 psuedoclass-plans.md | 63 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 63 insertions(+)
 create mode 100644 psuedoclass-plans.md

diff --git a/psuedoclass-plans.md b/psuedoclass-plans.md
new file mode 100644
index 0000000..3b979f4
--- /dev/null
+++ b/psuedoclass-plans.md
@@ -0,0 +1,63 @@
+Currently Stylish Haskell lacks support for psuedo-classes/elements, which is a significant shortcoming. So what would it take?
+
+To start the following psuedoclasses need to be defined by the caller (mostly in implementing forms), either before or after the CSS is evaluated:
+
+* active
+* checked
+* default
+* defined -- selects builtin elements or "custom elements".
+* disabled
+* empty -- since I do not directly reference children, just parents.
+* enabled
+* focus
+* focus-within
+* hover
+* indeterminate
+* in-range
+* invalid
+* optional
+* out-of-range
+* read-only
+* read-write
+* required
+* target
+* valid
+* visited
+
+The pseudoelements meanwhile need to be applied after selection:
+* before
+* after
+* cue
+* first-letter
+* first-line
+* selection
+
+:link, :lang(), & :root/:scope can compile down to existing selectorFuncs. 
+
+Furthermore there's those psuedoclasses which tests the position of the element in the list.
+* first-child
+* first-of-type
+* last-child
+* last-of-type
+* only-child
+* only-of-type
+* :nth-child()
+* :nth-last-child()
+* :nth-last-of-type()
+* :nth-of-type()
+:first-child, :first-of-type, :nth-child(), :nth-of-type() can easily be tested by traversing prev pointers.
+But for only/last psuedoclasses I'd need to store additional counts.
+
+Furthermore there's functional psuedoclasses which takes additional arguments:
+* :not(selector)
+* :where()/:is()
+I need to compile these down myself.
+
+I'd be tempted to output a structure mapping psuedoclasses to style objects,
+after having compiled some of those as part of the selector itself.
+Though the issue with that is that it prevents popular solutions for creating JavaScript-free webpages.
+
+I guess I'd add a psuedoclass field to elements, and there's already a hook
+where embedders can decide how they want to deal with it.
+
+Sorry this document is a bit rambly!
-- 
2.30.2