| 1 | #!/usr/bin/env bash
|
| 2 | #
|
| 3 | # Usage:
|
| 4 | # data_lang/htm8-test.sh
|
| 5 | #
|
| 6 | # TODO:
|
| 7 | # - Validate()
|
| 8 | # - add LEX_QUOTED_VALUES, along with counter for it
|
| 9 | # - and then re-run all the tests - make sure they pass
|
| 10 | # - site oils.pub, site oilshell.org
|
| 11 | # - XML on my machine - turn that in to 'WILD' corpus for HTML/XML?
|
| 12 | # - Rename to data_lang/htm8.py
|
| 13 | # - it has NO_SPECIAL_TAGS mode for XML
|
| 14 | # - put iterators at a higher level in doctools/ ?
|
| 15 | #
|
| 16 | # - statically type it
|
| 17 | # - revive pyannotate
|
| 18 | # - translate to C++
|
| 19 | # - how to handle the regexes in the lexer? Port to re2c directly?
|
| 20 | # - for find(), do we need a C++ primitive for it?
|
| 21 | # - no allocation for TagName()
|
| 22 | # - ASDL file for Tok.Foo?
|
| 23 | # - refactor TagName() API - remove it from the TagLexer?
|
| 24 | # - that is really the AttrLexer()
|
| 25 | #
|
| 26 | # Not working yet:
|
| 27 | # - understanding all entities &zz;
|
| 28 | # - there are over 2000 of them, not sure I want to build them all into the Oils binaries
|
| 29 | # - capital letters <TR/> - I guess we can normalize the case
|
| 30 | #
|
| 31 | # Features:
|
| 32 | # - Are there special rules for <svg> and <math>?
|
| 33 | # - Do we need to know about <textarea> <pre>? Those don't have the same
|
| 34 | # whitespace rules
|
| 35 | #
|
| 36 | # YSH API
|
| 37 | # - Generating HTML/HTM8 is much more common than parsing it
|
| 38 | # - although maybe we can do RemoveComments as a demo?
|
| 39 | # - that is the lowest level "sed" model
|
| 40 | # - For parsing, a minimum idea is:
|
| 41 | # - lexer-based algorithms for query by tag, class name, and id
|
| 42 | # - and then toTree() - this is a DOM
|
| 43 | # - .tag and .attrs?
|
| 44 | # - .innerHTML() and .outerHTML() perhaps
|
| 45 | # - rewrite ul-table in that?
|
| 46 | # - does that mean you mutate it, or construct text?
|
| 47 | # - I think you can set the innerHTML probably
|
| 48 | #
|
| 49 | # - Testing of html.ysh aka htm8.ysh in the stdlib
|
| 50 | #
|
| 51 | # Cases:
|
| 52 | # html 'hello <b>world</b>'
|
| 53 | # html "hello <b>$name</b>"html
|
| 54 | # html ["hello <b>$name</b>"] # hm this isn't bad, it's an unevaluated expression?
|
| 55 | # commonmark 'hello **world**'
|
| 56 | # md 'hello **world**'
|
| 57 | # md ['hello **$escape**'] ? We don't have a good escaping algorithm
|
| 58 |
|
| 59 |
|
| 60 | REPO_ROOT=$(cd "$(dirname $0)/.."; pwd)
|
| 61 |
|
| 62 | # Special case: we need $REPO_ROOT
|
| 63 | : ${LIB_OSH=$REPO_ROOT/stdlib/osh}
|
| 64 | source $LIB_OSH/bash-strict.sh
|
| 65 | source $LIB_OSH/task-five.sh
|
| 66 |
|
| 67 | # parse with lazylex/html.py, or data_lang/htm8.py
|
| 68 |
|
| 69 | site-files() {
|
| 70 | #find ../../oilshell/oilshell.org__deploy -name '*.html'
|
| 71 |
|
| 72 | # omit all the _ files
|
| 73 | git ls-files | grep '\.html$'
|
| 74 | }
|
| 75 |
|
| 76 | # Issues with lazylex/html.py
|
| 77 | #
|
| 78 | # - Token ID is annoying to express in Python
|
| 79 | # - re.DOTALL for newlines
|
| 80 | # - can we change that with [.\n]*?
|
| 81 | # - nongreedy match for --> and ?>
|
| 82 |
|
| 83 | htm8-tool() {
|
| 84 | PYTHONPATH="$REPO_ROOT:$REPO_ROOT/vendor" \
|
| 85 | $REPO_ROOT/lazylex/html.py "$@"
|
| 86 | }
|
| 87 |
|
| 88 | test-well-formed() {
|
| 89 | cat >_tmp/bad.html <<EOF
|
| 90 | unfinished <!--
|
| 91 | hi && bye
|
| 92 | EOF
|
| 93 | echo '_tmp/bad.html' | htm8-tool well-formed
|
| 94 | }
|
| 95 |
|
| 96 | # site errors
|
| 97 | #
|
| 98 | # Error in 'release/0.7.pre5/doc/osh-quick-ref.html': (LexError '&& or ||</h4>\n<!-- 2')
|
| 99 | # Error in 'src/data/symbol.html': (LexError "&& mangle[0]!='E' &&")
|
| 100 | # 5833374 tokens in 4710 files
|
| 101 | #
|
| 102 | # The second is the "Woboq" browser, which has CDATA
|
| 103 | # Ah I wonder if we need that.
|
| 104 |
|
| 105 | # Takes ~13 seconds
|
| 106 | test-site() {
|
| 107 | local new_site=${1:-}
|
| 108 |
|
| 109 | # TODO:
|
| 110 | # - test that the top level lexes
|
| 111 | # - test that each tag lexers
|
| 112 | # - test that each quoted attribute lexes
|
| 113 | # - test that tags are balanced
|
| 114 |
|
| 115 | local dir
|
| 116 | local action
|
| 117 | if test -n "$new_site"; then
|
| 118 | dir='../oils.pub__deploy'
|
| 119 | action='parse-htm8'
|
| 120 | else
|
| 121 | dir='../../oilshell/oilshell.org__deploy'
|
| 122 | action='lex-htm8'
|
| 123 | fi
|
| 124 |
|
| 125 | pushd $dir
|
| 126 |
|
| 127 | # Too many files
|
| 128 | # site-files | xargs wc -l | grep total
|
| 129 |
|
| 130 | # Not using xargs
|
| 131 | time site-files | $REPO_ROOT/$0 htm8-tool $action
|
| 132 |
|
| 133 | popd
|
| 134 | }
|
| 135 |
|
| 136 | readonly SOIL_ID=8917
|
| 137 | readonly WWZ_DIR=_tmp/$SOIL_ID
|
| 138 |
|
| 139 | sync-wwz() {
|
| 140 | mkdir -p $WWZ_DIR
|
| 141 | rsync --archive --verbose \
|
| 142 | op.oilshell.org:op.oilshell.org/uuu/github-jobs/$SOIL_ID/ $WWZ_DIR/
|
| 143 | }
|
| 144 |
|
| 145 | extract-wwz() {
|
| 146 | pushd $WWZ_DIR
|
| 147 | for z in *.wwz; do
|
| 148 | local name=$(basename $z .wwz)
|
| 149 |
|
| 150 | mkdir -p $name
|
| 151 | pushd $name >/dev/null
|
| 152 |
|
| 153 | echo $name
|
| 154 | unzip ../$z
|
| 155 |
|
| 156 | popd >/dev/null
|
| 157 | done
|
| 158 | popd
|
| 159 | }
|
| 160 |
|
| 161 | tree-wwz() {
|
| 162 | tree $WWZ_DIR
|
| 163 | }
|
| 164 |
|
| 165 | test-wwz() {
|
| 166 | pushd $WWZ_DIR
|
| 167 |
|
| 168 | time find . -name '*.html' | $REPO_ROOT/$0 htm8-tool parse-htm8
|
| 169 |
|
| 170 | popd
|
| 171 | }
|
| 172 |
|
| 173 | find-xml() {
|
| 174 | time find ~ -iname '*.xml' | tee _tmp/xml-files.txt
|
| 175 | }
|
| 176 |
|
| 177 | test-other-xml() {
|
| 178 | # problem with &ent1;
|
| 179 | # CDATA support! haha OK
|
| 180 | time cat _tmp/xml-files.txt | $REPO_ROOT/$0 htm8-tool parse-xml
|
| 181 | }
|
| 182 |
|
| 183 | test-repo-xml() {
|
| 184 | # OK these parse
|
| 185 | time find . -name '_chroot' -a -prune -o -name '*.xml' -a -print \
|
| 186 | | $REPO_ROOT/$0 htm8-tool parse-xml
|
| 187 | }
|
| 188 |
|
| 189 | test-repo-html() {
|
| 190 | time find . -name '*.html' | $REPO_ROOT/$0 htm8-tool parse-htm8
|
| 191 | }
|
| 192 |
|
| 193 | test-docs() {
|
| 194 | time find _release/VERSION -name '*.html' | $REPO_ROOT/$0 htm8-tool parse-htm8
|
| 195 | }
|
| 196 |
|
| 197 | soil-run() {
|
| 198 | test-docs
|
| 199 | }
|
| 200 |
|
| 201 | # OK we have to skip the <script> tag! And <style>
|
| 202 | #
|
| 203 | # document.location = '#' + params.join('&');
|
| 204 | # gUrlHash = new UrlHash(location.hash);
|
| 205 | #
|
| 206 | # I think textarea we don't though?
|
| 207 |
|
| 208 |
|
| 209 | task-five "$@"
|
| 210 | exit
|
| 211 |
|
| 212 |
|
| 213 | echo '
|
| 214 | In HTML5, instead of
|
| 215 | <script>
|
| 216 | <![CDATA[
|
| 217 | if (x < y) { ... }
|
| 218 | ]]>
|
| 219 | </script>
|
| 220 |
|
| 221 | You can write
|
| 222 |
|
| 223 | <script>
|
| 224 | if (x < y) { ... }
|
| 225 | </script>
|
| 226 |
|
| 227 | <script> <style> <textarea>
|
| 228 |
|
| 229 | These have special escaping rules. I guess we just do NOT lex them at all?
|
| 230 | We can totally SKIP them.
|
| 231 |
|
| 232 | CDATA vs. RCDATA
|
| 233 |
|
| 234 | <textarea>
|
| 235 | <p> <!-- This will show as: <p> -->
|
| 236 | & <!-- This will show as: & -->
|
| 237 | </textarea>
|
| 238 |
|
| 239 | <script>
|
| 240 | <p> <!-- This will show literally as: <p> -->
|
| 241 | & <!-- This will show literally as: & -->
|
| 242 | </script>
|
| 243 |
|
| 244 | The main practical difference is that RCDATA processes HTML entities while
|
| 245 | CDATA treats them as literal text. Both modes ignore HTML tags (treating them
|
| 246 | as plain text) except for their own closing tag. '
|
| 247 | '
|
| 248 |
|