# along with this program. If not, see <http://www.gnu.org/licenses/>.
-# This file implements a parser for html snippets, meant to be used by a
+# This file implements a thorough parser for html5, meant to be used by a
# WYSIWYG editor.
# The implementation is a pretty direct implementation of the parsing algorithm
# described here:
-# http://www.w3.org/TR/html5/syntax.html#preprocessing-the-input-stream
#
-# Deviations from that spec:
+# http://www.w3.org/TR/html5/syntax.html
#
-# Purposeful: search this file for "WHATWG"
+# except for some places marked "WHATWG" that are implemented as described here:
#
-# Not finished yet: search this file for "fixfull", "TODO" and "FIXME"
+# https://html.spec.whatwg.org/multipage/syntax.html
+#
+# This code passes all of the tests in the .dat files at:
+#
+# https://github.com/JasonWoof/html5lib-tests/tree/patch-1/tree-construction
+
+
+##################################
+## how to use this code
+##################################
+#
+# See README.md for how to run this file in the browser or in node.js.
+#
+# This file exports a single useful function: parse_tml, and some constants
+# (see the bottom of this file for those.)
+#
+# Call it like this:
+#
+# wheic.parse_html("<p><b>hi</p>")
+#
+# Or, if you don't want <html><head><body>/etc, do this:
+#
+# wheic.parse_html("<p><b>hi</p>", {fragment: "body"})
+#
+# return value is an array of Nodes, see "class Node" below.
+
+# This code is a work in progress, eg try search this file for "fixfull",
+# "TODO" and "FIXME"
-# stacks/lists
+# Notes: stacks/lists
#
-# the spec uses a many different words do indicate which ends of lists/stacks
-# they are talking about (and relative movement within the lists/stacks). This
-# section splains. I'm implementing "lists" (afe and open_els) the same way
-# (both as stacks)
+# Jason was frequently confused by the terminology used to refer to different
+# parts of the stacks and lists in the spec, so he made this chart to help keep
+# his head straight:
#
# stacks grow downward (current element is index=0)
#
# example: open_els = [a, b, c, d, e, f, g]
#
-# "grows downwards" means it's visualized like this: (index: el, names)
+# "grows downwards" means it's visualized like this: (index: el "names")
#
# 6: g "start of the list", "topmost", "first"
# 5: f
# 1: b
# 0: a "end of the list", "current node", "bottommost", "last"
-
-# browser
-# note: to get this to run outside a browser, you'll have to write a native
-# implementation of decode_named_char_ref()
-unless module?.exports?
+if (typeof module) isnt 'undefined' and module.exports?
+ context = 'module'
+ exports = module.exports
+else
+ context = 'browser'
window.wheic = {}
- module = exports: window.wheic
+ exports = window.wheic
from_code_point = (x) ->
if String.fromCodePoint?
QUIRKS_LIMITED = 2
QUIRKS_YES = 3
+# queue up debug logs, so eg they can be shown only for tests that fail
g_debug_log = []
debug_log_reset = ->
g_debug_log = []
+ return
debug_log = (str) ->
g_debug_log.push str
+ return
debug_log_each = (cb) ->
for str in g_debug_log
cb str
+ return
prev_node_id = 0
class Node
@token.flag 'did_self_close', true
else
@flag 'did_self_close', true
+ return
flag: (key, value = null) ->
if value?
@flags[key] = value
else
return @flags[key]
- serialize: (shallow = false, show_ids = false) -> # for unit tests
- ret = ''
- switch @type
- when TYPE_TAG
- ret += 'tag:'
- ret += JSON.stringify @name
- ret += ','
- if show_ids
- ret += "##{@id},"
- if shallow
- break
- attr_keys = []
- for k of @attrs
- attr_keys.push k
- attr_keys.sort()
- ret += '{'
- sep = ''
- for k in attr_keys
- ret += sep
- sep = ','
- ret += "#{JSON.stringify k}:#{JSON.stringify @attrs[k]}"
- ret += '},['
- sep = ''
- for c in @children
- ret += sep
- sep = ','
- ret += c.serialize shallow, show_ids
- ret += ']'
- when TYPE_TEXT
- ret += 'text:'
- ret += JSON.stringify @text
- when TYPE_COMMENT
- ret += 'comment:'
- ret += JSON.stringify @text
- when TYPE_DOCTYPE
- ret += "doctype:#{@name},#{JSON.stringify(@public_identifier ? '')},#{JSON.stringify(@system_identifier ? '')}"
- when TYPE_AFE_MARKER
- ret += 'marker'
- when TYPE_AAA_BOOKMARK
- ret += 'aaa_bookmark'
- else
- ret += 'unknown:'
- console.log "unknown: #{JSON.stringify @}" # backtrace is just as well
- return ret
+ return
# helpers: (only take args that are normally known when parser creates nodes)
new_open_tag = (name) ->
# decode_named_char_ref()
#
-# The list of named character references is _huge_ so ask the browser to decode
-# for us instead of wasting bandwidth/space on including the table here.
-#
-# Pass without the "&" but with the ";" examples:
-# for "&" pass "amp;"
-# for "′" pass "x2032;"
-g_dncr = {
- cache: {}
- textarea: document.createElement('textarea')
-}
-# TODO test this in IE8
+# The list of named character references is _huge_ so if we're running in a
+# browser, we get the browser to decode them, rather than increasing the code
+# size to include the table.
+if context is 'module'
+ _decode_named_char_ref = require './html5-named-entities.coffee'
+else
+ # TODO test this in IE8
+ decode_named_char_ref_el = document.createElement('textarea')
+ _decode_named_char_ref = (txt) ->
+ txt = "&#{txt};"
+ decode_named_char_ref_el.innerHTML = txt
+ decoded = decode_named_char_ref_el.value
+ return null if decoded is txt
+ return decoded
+# Pass the name of a named entity _that has a terminating semicolon_
+# Entities without terminating semicolons should use legacy_char_refs[]
+# Do not include the "&" or ";" in your argument, eg pass "alpha"
+decode_named_char_ref_cache = {}
decode_named_char_ref = (txt) ->
- txt = "&#{txt}"
- decoded = g_dncr.cache[txt]
+ decoded = decode_named_char_ref_cache[txt]
return decoded if decoded?
- g_dncr.textarea.innerHTML = txt
- decoded = g_dncr.textarea.value
- return null if decoded is txt
- return g_dncr.cache[txt] = decoded
+ decoded = _decode_named_char_ref txt
+ return decode_named_char_ref_cache[txt] = decoded
-parse_html = (args) ->
+parse_html = (args_html, args = {}) ->
txt = null
cur = null # index of next char in txt to be parsed
# declare doc and tokenizer variables so they're in scope below
stop_parsing = ->
flag_parsing = false
+ return
parse_error = ->
if args.error_cb?
args.error_cb cur
else
console.log "Parse error at character #{cur} of #{txt.length}"
+ return
# http://www.w3.org/TR/html5/syntax.html#push-onto-the-list-of-active-formatting-elements
# "Noah's Ark clause" but with three
afe.splice i, 1
break
afe.unshift new_el
+ return
+
afe_push_marker = ->
afe.unshift new_afe_marker()
+ return
# the functions below impliment the Tree Contstruction algorithm
# http://www.w3.org/TR/html5/syntax.html#tree-construction
loop
if node_i is open_els.length - 1
last = true
- # fixfull (fragment case)
-
+ if flag_fragment_parsing
+ node = context_element
# 4. If node is a select element, run these substeps:
if node.name is 'select' and node.namespace is NS_HTML
# 1. If last is true, jump to the step below labeled done.
node_i += 1
node = open_els[node_i]
# 19. Return to the step labeled loop.
+ return
# 8.2.3.2
afe[i] = el
break if i is 0
i -= 1 # Advance
+ return
# http://www.w3.org/TR/html5/syntax.html#adoption-agency-algorithm
# adoption agency algorithm
# http://www.w3.org/TR/html5/syntax.html#misnested-tags:-b-p-/b-/p
# http://www.w3.org/TR/html5/syntax.html#unclosed-formatting-elements
adoption_agency = (subject) ->
- debug_log "adoption_agency()"
- debug_log "tree: #{serialize_els doc.children, false, true}"
- debug_log "open_els: #{serialize_els open_els, true, true}"
- debug_log "afe: #{serialize_els afe, true, true}"
# this block implements tha W3C spec
# # 1. If the current node is an HTML element whose tag name is subject,
# # then run these substeps:
# if t is el
# afe.splice i, 1
# break
-# debug_log "aaa: starting off with subject on top of stack, exiting"
# return
# WHATWG: https://html.spec.whatwg.org/multipage/syntax.html#adoption-agency-algorithm
# If the current node is an HTML element whose tag name is subject, and
# then pop the current node off the stack of open elements, and abort
# these steps.
if open_els[0].name is subject and open_els[0].namespace is NS_HTML
- debug_log "aaa: starting off with subject on top of stack, exiting"
# remove it from the list of active formatting elements (if found)
in_afe = false
for el, i in afe
in_afe = true
break
unless in_afe
- debug_log "aaa: ...and not in afe, aaa done"
open_els.shift()
return
# fall through
# If there is no such element, then abort these steps and instead
# act as described in the "any other end tag" entry above.
if fe is null
- debug_log "aaa: fe not found in afe"
in_body_any_other_end_tag subject
return
# 6. If formatting element is not in the stack of open elements,
in_open_els = true
break
unless in_open_els
- debug_log "aaa: fe not found in open_els"
parse_error()
# "remove it from the list" must mean afe, since it's not in open_els
afe.splice fe_of_afe, 1
# the element is not in scope, then this is a parse error; abort
# these steps.
unless el_is_in_scope fe
- debug_log "aaa: fe not in scope"
parse_error()
return
# 8. If formatting element is not the current node, this is a parse
# formatting element from the list of active formatting elements,
# and finally abort these steps.
if fb is null
- debug_log "aaa: no fb"
loop
t = open_els.shift()
if t is fe
node_next = open_els[i + 1]
break
node = node_next ? node_above
- debug_log "inner loop #{inner}"
- debug_log "tree: #{serialize_els doc.children, false, true}"
- debug_log "open_els: #{serialize_els open_els, true, true}"
- debug_log "afe: #{serialize_els afe, true, true}"
- debug_log "ca: #{ca.name}##{ca.id} children: #{serialize_els ca.children, true, true}"
- debug_log "fe: #{fe.name}##{fe.id} children: #{serialize_els fe.children, true, true}"
- debug_log "fb: #{fb.name}##{fb.id} children: #{serialize_els fb.children, true, true}"
- debug_log "node: #{node.serialize true, true}"
# TODO make sure node_above gets re-set if/when node is removed from open_els
# 4. If node is formatting element, then go to the next step in
# the overall algorithm.
if node is fe
break
- debug_log "the meat"
# 5. If inner loop counter is greater than three and node is in
# the list of active formatting elements, then remove node from
# the list of active formatting elements.
if t is node
if inner > 3
afe.splice i, 1
- debug_log "max out inner"
else
node_in_afe = true
- debug_log "in afe"
break
# 6. If node is not in the list of active formatting elements,
# then remove node from the stack of open elements and then go
# back to the step labeled inner loop.
unless node_in_afe
- debug_log "not in afe"
for t, i in open_els
if t is node
node_above = open_els[i + 1]
open_els.splice i, 1
break
continue
- debug_log "the bones"
# 7. create an element for the token for which the element node
# was created, in the HTML namespace, with common ancestor as
# the intended parent; replace the entry for node in the list
for t, i in afe
if t is node
afe[i] = new_node
- debug_log "replaced in afe"
break
for t, i in open_els
if t is node
node_above = open_els[i + 1]
open_els[i] = new_node
- debug_log "replaced in open_els"
break
node = new_node
# 8. If last node is furthest block, then move the
for t, i in afe
if t is bookmark
afe.splice i, 1
- debug_log "removed bookmark"
break
for t, i in afe
if t is node
# "after" means lower
afe.splice i, 0, bookmark # "after as <-
- debug_log "placed bookmark after node"
- debug_log "node: #{node.id} afe: #{serialize_els afe, true, true}"
break
# 9. Insert last node into node, first removing it from its
# previous parent node if any.
if last_node.parent?
- debug_log "last_node has parent"
for c, i in last_node.parent.children
if c is last_node
- debug_log "removing last_node from parent"
last_node.parent.children.splice i, 1
break
node.children.push last_node
last_node.parent = node
# 10. Let last node be node.
last_node = node
- debug_log "at last"
# 11. Return to the step labeled inner loop.
# 14. Insert whatever last node ended up being in the previous step
# at the appropriate place for inserting a node, but using common
# * last_node is fb
# * last_node is still in the tree (not a duplicate)
if last_node.parent?
- debug_log "FEFIRST? last_node has parent"
for c, i in last_node.parent.children
if c is last_node
- debug_log "removing last_node from parent"
last_node.parent.children.splice i, 1
break
-
- debug_log "after aaa inner loop"
- debug_log "ca: #{ca.name}##{ca.id} children: #{serialize_els ca.children, true, true}"
- debug_log "fe: #{fe.name}##{fe.id} children: #{serialize_els fe.children, true, true}"
- debug_log "fb: #{fb.name}##{fb.id} children: #{serialize_els fb.children, true, true}"
- debug_log "last_node: #{last_node.name}##{last_node.id} children: #{serialize_els last_node.children, true, true}"
- debug_log "tree: #{serialize_els doc.children, false, true}"
-
- debug_log "insert"
-
-
# can't use standard insert token thing, because it's already in
# open_els and must stay at it's current position in open_els
dest = adjusted_insertion_location ca
dest[0].children.splice dest[1], 0, last_node
last_node.parent = dest[0]
-
-
- debug_log "ca: #{ca.name}##{ca.id} children: #{serialize_els ca.children, true, true}"
- debug_log "fe: #{fe.name}##{fe.id} children: #{serialize_els fe.children, true, true}"
- debug_log "fb: #{fb.name}##{fb.id} children: #{serialize_els fb.children, true, true}"
- debug_log "last_node: #{last_node.name}##{last_node.id} children: #{serialize_els last_node.children, true, true}"
- debug_log "tree: #{serialize_els doc.children, false, true}"
-
# 15. Create an element for the token for which formatting element
# was created, in the HTML namespace, with furthest block as the
# intended parent.
open_els.splice i, 0, new_element
break
# 20. Jump back to the step labeled outer loop.
- debug_log "done wrapping fb's children. new_element: #{new_element.name}##{new_element.id}"
- debug_log "tree: #{serialize_els doc.children, false, true}"
- debug_log "open_els: #{serialize_els open_els, true, true}"
- debug_log "afe: #{serialize_els afe, true, true}"
- debug_log "AAA DONE"
+ return
# http://www.w3.org/TR/html5/syntax.html#close-a-p-element
close_p_element = ->
el = open_els.shift()
if el.name is 'p' and el.namespace is NS_HTML
return
+ return
close_p_if_in_button_scope = ->
if is_in_button_scope 'p', NS_HTML
close_p_element()
+ return
# http://www.w3.org/TR/html5/syntax.html#insert-a-character
# aka insert_a_character = (t) ->
prev.text += t.text
return
dest[0].children.splice dest[1], 0, t
-
+ return
# 8.2.5 http://www.w3.org/TR/html5/syntax.html#tree-construction
process_token = (t) ->
return el
# http://www.w3.org/TR/html5/syntax.html#insert-an-html-element
insert_html_element = (token) ->
- insert_foreign_element token, NS_HTML
+ return insert_foreign_element token, NS_HTML
# http://www.w3.org/TR/html5/syntax.html#insert-a-comment
# position should be [node, index_within_children]
insert_comment = (t, position = null) ->
position ?= adjusted_insertion_location()
position[0].children.splice position[1], 0, t
+ return
# 8.2.5.2
# http://www.w3.org/TR/html5/syntax.html#generic-raw-text-element-parsing-algorithm
tok_state = tok_state_rawtext
original_ins_mode = ins_mode
ins_mode = ins_mode_text
+ return
parse_generic_rcdata_text = (t) ->
insert_html_element t
tok_state = tok_state_rcdata
original_ins_mode = ins_mode
ins_mode = ins_mode_text
+ return
# 8.2.5.3 http://www.w3.org/TR/html5/syntax.html#closing-elements-that-have-implied-end-tags
# http://www.w3.org/TR/html5/syntax.html#generate-implied-end-tags
generate_implied_end_tags = (except = null) ->
while end_tag_implied[open_els[0].name] is open_els[0].namespace and open_els[0].name isnt except
open_els.shift()
+ return
# 8.2.5.4 The rules for parsing tokens in HTML content
# http://www.w3.org/TR/html5/syntax.html#parsing-main-inhtml
if t.type is TYPE_START_TAG and t.name is 'html'
el = token_to_element t, NS_HTML, doc
doc.children.push el
+ el.document = doc
open_els.unshift(el)
# fixfull (big paragraph in spec about manifest, fragment, urls, etc)
ins_mode = ins_mode_before_head
# Anything else
el = token_to_element new_open_tag('html'), NS_HTML, doc
doc.children.push el
- el.parent = doc
+ el.document = doc
open_els.unshift el
# ?fixfull browsing context
ins_mode = ins_mode_before_head
head_element_pointer = el
ins_mode = ins_mode_in_head
process_token t
+ return
# 8.2.5.4.4 http://www.w3.org/TR/html5/syntax.html#parsing-main-inhead
ins_mode_in_head_else = (t) -> # factored out for same-as-spec flow control
open_els.shift() # spec says this will be a 'head' node
ins_mode = ins_mode_after_head
process_token t
+ return
ins_mode_in_head = (t) ->
if t.type is TYPE_TEXT and (t.text is "\t" or t.text is "\n" or t.text is "\u000c" or t.text is ' ')
insert_character t
parse_error()
return
ins_mode_in_head_else t
+ return
# 8.2.5.4.5 http://www.w3.org/TR/html5/syntax.html#parsing-main-inheadnoscript
ins_mode_in_head_noscript_else = (t) ->
open_els.shift()
ins_mode = ins_mode_in_head
process_token t
+ return
ins_mode_in_head_noscript = (t) ->
if t.type is TYPE_DOCTYPE
parse_error()
ins_mode_in_head_noscript_else t
return
-
-
# 8.2.5.4.6 http://www.w3.org/TR/html5/syntax.html#the-after-head-insertion-mode
ins_mode_after_head_else = (t) ->
body_tok = new_open_tag 'body'
if el is head_element_pointer
open_els.splice i, 1
return
- console.log "warning: 23904 couldn't find head element in open_els"
return
if t.type is TYPE_END_TAG and t.name is 'template'
ins_mode_in_head t
return
# Anything else
ins_mode_after_head_else t
+ return
# 8.2.5.4.7 http://www.w3.org/TR/html5/syntax.html#parsing-main-inbody
in_body_any_other_end_tag = (name) -> # factored out because adoption agency calls it
return
if t.type is TYPE_END_TAG and t.name is 'br'
parse_error()
- t.type = TYPE_START_TAG
+ # W3C: t.type = TYPE_START_TAG
+ t = new_open_tag 'br' # WHATWG
# fall through
if t.type is TYPE_START_TAG and (t.name is 'area' or t.name is 'br' or t.name is 'embed' or t.name is 'img' or t.name is 'keygen' or t.name is 'wbr')
reconstruct_afe()
open_els.shift()
ins_mode = original_ins_mode
return
- console.log 'warning: end of ins_mode_text reached'
+ return
# the functions below implement the tokenizer stats described here:
# http://www.w3.org/TR/html5/syntax.html#tokenization
ins_mode_in_body t
else
ins_mode_in_table_else t
+ return
# 8.2.5.4.10 http://www.w3.org/TR/html5/syntax.html#parsing-main-intabletext
pending_table_character_tokens = []
ins_mode = original_ins_mode
process_token t
+ return
# 8.2.5.4.11 http://www.w3.org/TR/html5/syntax.html#parsing-main-incaption
ins_mode_in_caption = (t) ->
return
# Anything else
ins_mode_in_body t
+ return
# 8.2.5.4.12 http://www.w3.org/TR/html5/syntax.html#parsing-main-incolgroup
ins_mode_in_column_group = (t) ->
return
# Anything else
ins_mode_in_table t
+ return
# 8.2.5.4.14 http://www.w3.org/TR/html5/syntax.html#parsing-main-intr
ins_mode_in_row = (t) ->
return
# Anything else
ins_mode_in_table t
+ return
# http://www.w3.org/TR/html5/syntax.html#close-the-cell
close_the_cell = ->
break
clear_afe_to_marker()
ins_mode = ins_mode_in_row
+ return
# 8.2.5.4.15 http://www.w3.org/TR/html5/syntax.html#parsing-main-intd
ins_mode_in_cell = (t) ->
return
# Anything Else
ins_mode_in_body t
+ return
# 8.2.5.4.16 http://www.w3.org/TR/html5/syntax.html#parsing-main-inselect
ins_mode_in_select = (t) ->
template_ins_modes.shift()
reset_ins_mode()
process_token t
+ return
# 8.2.5.4.19 http://www.w3.org/TR/html5/syntax.html#parsing-main-afterbody
ins_mode_after_body = (t) ->
parse_error()
ins_mode = ins_mode_in_body
process_token t
+ return
# 8.2.5.4.20 http://www.w3.org/TR/html5/syntax.html#parsing-main-inframeset
ins_mode_in_frameset = (t) ->
if node.namespace is NS_HTML
break
ins_mode t # explicitly call HTML insertion mode
+ return
# 8.2.4.1 http://www.w3.org/TR/html5/syntax.html#data-state
tok_state = tok_state_tag_open
when "\u0000"
parse_error()
- return new_text_node "\ufffd"
+ return new_text_node c
when '' # EOF
return new_eof_token()
else
# http://www.w3.org/TR/html5/syntax.html#appropriate-end-tag-token
is_appropriate_end_tag = (t) ->
- # spec says to check against "the tag name of the last start tag to
- # have been emitted from this tokenizer", but this is only called from
- # the various "raw" states, so it's hopefully ok to assume that
- # open_els[0].name will work instead TODO: verify this after the script
- # data states are implemented
- debug_log "#{t.type}, #{t.name} open_els: #{serialize_els open_els, true, true}"
+ # fixfull: this assumes that open_els[0].name is "the tag name of the last
+ # start tag to have been emitted from this tokenizer"
return t.type is TYPE_END_TAG and t.name is open_els[0].name
# 8.2.4.13 http://www.w3.org/TR/html5/syntax.html#rcdata-end-tag-name-state
# Anything else
tok_cur_tag.attrs_a.unshift [c, '']
tok_state = tok_state_attribute_name
+ return
# 8.2.4.37 http://www.w3.org/TR/html5/syntax.html#before-attribute-value-state
tok_state_before_attribute_value = ->
else
val = txt.substr cur, (next_gt - cur)
cur = next_gt + 3
+ val = val.replace(new RegExp("\u0000", 'g'), "\ufffd")
if val.length > 0
return new_character_token val # fixfull split
return null
# exit early, because parse_error() below needs at least one alnum
return '&'
if txt.charAt(cur + i) is ';'
- i += 1 # include ';' terminator in value
decoded = decode_named_char_ref txt.substr(cur, i)
+ i += 1 # scan past the ';' (after, so we dno't pass it to decode)
if decoded?
cur += i
return decoded
- parse_error()
- return '&'
- else
- # no ';' terminator (only legacy char refs)
- max = i
- for i in [2..max] # no prefix matches, so ok to check shortest first
- c = legacy_char_refs[txt.substr(cur, i)]
- if c?
- if in_attr
- if txt.charAt(cur + i) is '='
- # "because some legacy user agents will
- # misinterpret the markup in those cases"
- parse_error()
- return '&'
- if alnum.indexOf(txt.charAt(cur + i)) > -1
- # this makes attributes forgiving about url args
- return '&'
- # ok, and besides the weird exceptions for attributes...
- # return the matching char
- cur += i # consume entity chars
- parse_error() # because no terminating ";"
- return c
- parse_error()
- return '&'
+ # else FALL THROUGH (check for match without last char(s) or ";")
+ # no ';' terminator (only legacy char refs)
+ max = i
+ for i in [2..max] # no prefix matches, so ok to check shortest first
+ c = legacy_char_refs[txt.substr(cur, i)]
+ if c?
+ if in_attr
+ if txt.charAt(cur + i) is '='
+ # "because some legacy user agents will
+ # misinterpret the markup in those cases"
+ parse_error()
+ return '&'
+ if alnum.indexOf(txt.charAt(cur + i)) > -1
+ # this makes attributes forgiving about url args
+ return '&'
+ # ok, and besides the weird exceptions for attributes...
+ # return the matching char
+ cur += i # consume entity chars
+ parse_error() # because no terminating ";"
+ return c
+ parse_error()
+ return '&'
return # never reached
eat_next_token_if_newline = ->
# tree constructor initialization
# see comments on TYPE_TAG/etc for the structure of this data
- txt = args.html
+ txt = args_html
cur = 0
- doc = new Node TYPE_TAG, name: 'html', namespace: NS_HTML
+ doc = new Node TYPE_TAG, name: 'document', namespace: NS_HTML
doc.flag 'quirks mode', QUIRKS_NO # TODO bugreport spec for not specifying this
+ fragment_root = null # fragment parsing algorithm returns children of this
open_els = []
afe = [] # active formatting elements
template_ins_modes = []
temporary_buffer = null
pending_table_character_tokens = []
head_element_pointer = null
- flag_fragment_parsing = false # parser originally created as part of the html fragment parsing algorithm (fragment case)
- context_element = null # FIXME initialize from args.fragment http://www.w3.org/TR/html5/syntax.html#parsing-html-fragments
+ flag_fragment_parsing = false
+ context_element = null
prev_node_id = 0 # just for debugging
# tokenizer initialization
tok_state = tok_state_data
- # text pre-processing
- # FIXME http://www.w3.org/TR/html5/syntax.html#preprocessing-the-input-stream
- txt = txt.replace(new RegExp("\u0000", 'g'), "\ufffd") # fixfull spec doesn't say this
- txt = txt.replace(new RegExp("\r\n", 'g'), "\n") # fixfull spec doesn't say this
- txt = txt.replace(new RegExp("\r", 'g'), "\n") # fixfull spec doesn't say this
+ parse_init = ->
+ # fragment parsing (text arg)
+ if args.fragment?
+ # this handles the fragment from the tests in the format described here:
+ # https://github.com/html5lib/html5lib-tests/blob/master/tree-construction/README.md
+ f = args.fragment
+ ns = NS_HTML
+ if f.substr(0, 5) is 'math '
+ f = f.substr 5
+ ns = NS_MATHML
+ else if f.substr(0, 4) is 'svg '
+ f = f.substr 4
+ ns = NS_SVG
+ t = new_open_tag f
+ context_element = token_to_element t, ns
+ context_element.document = new Node TYPE_TAG, name: 'document', namespace: NS_HTML
+ context_element.document.flag 'quirks mode', QUIRKS_NO
+ # fragment parsing (Node arg)
+ if args.context?
+ context_element = args.context
+
+ # http://www.w3.org/TR/html5/syntax.html#parsing-html-fragments
+ # fragment parsing algorithm
+ if context_element?
+ flag_fragment_parsing = true
+ doc = new Node TYPE_TAG, name: 'html', namespace: NS_HTML
+ # search up the tree from context, to try to find it's document,
+ # because this file only puts a "document" property on the root
+ # element.
+ old_doc = null
+ el = context_element
+ loop
+ if el.document?
+ old_doc = el.document
+ break
+ if el.parent
+ el = el.parent
+ else
+ break
+ if old_doc
+ doc.flag 'quirks mode', old_doc.flag 'quirks mode'
+ # set tok_state
+ if context_element.namespace is NS_HTML
+ switch context_element.name
+ when 'title', 'textarea'
+ tok_state = tok_state_rcdata
+ when 'style', 'xmp', 'iframe', 'noembed', 'noframes'
+ tok_state = tok_state_rawtext
+ when 'script'
+ tok_state = tok_state_script_data
+ when 'noscript'
+ if flag_scripting
+ tok_state = tok_state_rawtext
+ when 'plaintext'
+ tok_state = tok_state_plaintext
+ fragment_root = new Node TYPE_TAG, name: 'html', namespace: NS_HTML
+ doc.children.push fragment_root
+ fragment_root.document = doc
+ open_els = [fragment_root]
+ if context_element.name is 'template' and context_element.namespace is NS_HTML
+ template_ins_modes.unshift ins_mode_in_template
+ # fixfull create token for context (it should have it's original one already)
+ reset_ins_mode()
+ # set form_element pointer... in the foreign doc?!
+ el = context_element
+ loop
+ if el.name is 'form' and el.namespace is NS_HTML
+ form_element_pointer = el
+ break
+ if el.parent
+ el = el.parent
+ else
+ break
+
+ # text pre-processing
+ # FIXME check http://www.w3.org/TR/html5/syntax.html#preprocessing-the-input-stream
+ txt = txt.replace(new RegExp("\r\n", 'g'), "\n") # fixfull spec doesn't say this
+ txt = txt.replace(new RegExp("\r", 'g'), "\n") # fixfull spec doesn't say this
+
+ return
- if args.name is "webkit01.dat #12"
- console.log "hi"
- # proccess input
# http://www.w3.org/TR/html5/syntax.html#tree-construction
parse_main_loop = ->
while flag_parsing
if t?
process_token t
# fixfull parse error if has self-closing flag, but it wasn't acknolwedged
+ return
+ parse_init()
parse_main_loop()
+
+ if flag_fragment_parsing
+ return fragment_root.children
return doc.children
-serialize_els = (els, shallow, show_ids) ->
- serialized = ''
- sep = ''
- for t in els
- serialized += sep
- sep = ','
- serialized += t.serialize shallow, show_ids
- return serialized
-
-module.exports.parse_html = parse_html
-module.exports.debug_log_reset = debug_log_reset
-module.exports.debug_log_each = debug_log_each
-module.exports.TYPE_TAG = TYPE_TAG
-module.exports.TYPE_TEXT = TYPE_TEXT
-module.exports.TYPE_COMMENT = TYPE_COMMENT
-module.exports.TYPE_DOCTYPE = TYPE_DOCTYPE
-module.exports.NS_HTML = NS_HTML
-module.exports.NS_MATHML = NS_MATHML
-module.exports.NS_SVG = NS_SVG
-module.exports.QUIRKS_NO = QUIRKS_NO
-module.exports.QUIRKS_LIMITED = QUIRKS_LIMITED
-module.exports.QUIRKS_YES = QUIRKS_YES
+exports.parse_html = parse_html
+exports.debug_log_reset = debug_log_reset
+exports.debug_log_each = debug_log_each
+exports.TYPE_TAG = TYPE_TAG
+exports.TYPE_TEXT = TYPE_TEXT
+exports.TYPE_COMMENT = TYPE_COMMENT
+exports.TYPE_DOCTYPE = TYPE_DOCTYPE
+exports.NS_HTML = NS_HTML
+exports.NS_MATHML = NS_MATHML
+exports.NS_SVG = NS_SVG
+exports.QUIRKS_NO = QUIRKS_NO
+exports.QUIRKS_LIMITED = QUIRKS_LIMITED
+exports.QUIRKS_YES = QUIRKS_YES
Got questions, comments, patches, etc.? Contact Jason Woofenden