| 1 | #!/usr/bin/env bash
|
| 2 | #
|
| 3 | # Usage:
|
| 4 | # data_lang/htm8-test.sh
|
| 5 | #
|
| 6 | # TODO:
|
| 7 | # - Rename to DML8? Because it can handle XML
|
| 8 | # - CDATA in XML, which is not a script
|
| 9 | #
|
| 10 | # Operations / Levels:
|
| 11 | #
|
| 12 | # - Lexing
|
| 13 | # - lex-tags
|
| 14 | # - lex-attrs - validate all Start tags, all StartEnd tags
|
| 15 | # - lex-quoted-values - unescaping, etc.
|
| 16 | # - are there invalid entities?
|
| 17 | # - Parsing
|
| 18 | # - well-formed / tag balance check
|
| 19 | # - Schema
|
| 20 | # - not sure if we check the HTML schema or not - it might be too restrictive
|
| 21 |
|
| 22 | REPO_ROOT=$(cd "$(dirname $0)/.."; pwd)
|
| 23 |
|
| 24 | # Special case: we need $REPO_ROOT
|
| 25 | : ${LIB_OSH=$REPO_ROOT/stdlib/osh}
|
| 26 | source $LIB_OSH/bash-strict.sh
|
| 27 | source $LIB_OSH/task-five.sh
|
| 28 |
|
| 29 | # parse with lazylex/html.py, or data_lang/htm8.py
|
| 30 |
|
| 31 | site-files() {
|
| 32 | #find ../../oilshell/oilshell.org__deploy -name '*.html'
|
| 33 |
|
| 34 | # omit all the _ files
|
| 35 | git ls-files | grep '\.html$'
|
| 36 | }
|
| 37 |
|
| 38 | # Issues with lazylex/html.py
|
| 39 | #
|
| 40 | # - Token ID is annoying to express in Python
|
| 41 | # - re.DOTALL for newlines
|
| 42 | # - can we change that with [.\n]*?
|
| 43 | # - nongreedy match for --> and ?>
|
| 44 |
|
| 45 | ht8-tool() {
|
| 46 | PYTHONPATH="$REPO_ROOT:$REPO_ROOT/vendor" \
|
| 47 | $REPO_ROOT/lazylex/html.py "$@"
|
| 48 | }
|
| 49 |
|
| 50 | test-well-formed() {
|
| 51 | cat >_tmp/bad.html <<EOF
|
| 52 | unfinished <!--
|
| 53 | hi && bye
|
| 54 | EOF
|
| 55 | echo '_tmp/bad.html' | ht8-tool well-formed
|
| 56 | }
|
| 57 |
|
| 58 | # site errors
|
| 59 | #
|
| 60 | # Error in 'release/0.7.pre5/doc/osh-quick-ref.html': (LexError '&& or ||</h4>\n<!-- 2')
|
| 61 | # Error in 'src/data/symbol.html': (LexError "&& mangle[0]!='E' &&")
|
| 62 | # 5833374 tokens in 4710 files
|
| 63 | #
|
| 64 | # The second is the "Woboq" browser, which has CDATA
|
| 65 | # Ah I wonder if we need that.
|
| 66 |
|
| 67 | # Takes ~13 seconds
|
| 68 | test-site() {
|
| 69 | # TODO:
|
| 70 | # - test that the top level lexes
|
| 71 | # - test that each tag lexers
|
| 72 | # - test that each quoted attribute lexes
|
| 73 | # - test that tags are balanced
|
| 74 |
|
| 75 | pushd ../../oilshell/oilshell.org__deploy
|
| 76 |
|
| 77 | # Too many files
|
| 78 | # site-files | xargs wc -l | grep total
|
| 79 |
|
| 80 | # Not using xargs
|
| 81 | time site-files | $REPO_ROOT/$0 ht8-tool well-formed
|
| 82 |
|
| 83 | popd
|
| 84 | }
|
| 85 |
|
| 86 | readonly WWZ_DIR=_tmp/8899
|
| 87 |
|
| 88 | sync-wwz() {
|
| 89 | mkdir -p $WWZ_DIR
|
| 90 | rsync --archive --verbose \
|
| 91 | op.oilshell.org:op.oilshell.org/uuu/github-jobs/8899/ $WWZ_DIR/
|
| 92 | }
|
| 93 |
|
| 94 | extract-wwz() {
|
| 95 | pushd $WWZ_DIR
|
| 96 | for z in *.wwz; do
|
| 97 | local name=$(basename $z .wwz)
|
| 98 |
|
| 99 | mkdir -p $name
|
| 100 | pushd $name >/dev/null
|
| 101 |
|
| 102 | echo $name
|
| 103 | unzip ../$z
|
| 104 |
|
| 105 | popd >/dev/null
|
| 106 | done
|
| 107 | popd
|
| 108 | }
|
| 109 |
|
| 110 | tree-wwz() {
|
| 111 | tree $WWZ_DIR
|
| 112 | }
|
| 113 |
|
| 114 | test-wwz() {
|
| 115 | pushd $WWZ_DIR
|
| 116 |
|
| 117 | time find . -name '*.html' | $REPO_ROOT/$0 ht8-tool well-formed
|
| 118 |
|
| 119 | popd
|
| 120 | }
|
| 121 |
|
| 122 | find-xml() {
|
| 123 | time find ~ -iname '*.xml' | tee _tmp/xml-files.txt
|
| 124 | }
|
| 125 |
|
| 126 | test-other-xml() {
|
| 127 | # problem with &ent1;
|
| 128 | # CDATA support! haha OK
|
| 129 | time cat _tmp/xml-files.txt | $REPO_ROOT/$0 ht8-tool well-formed
|
| 130 | }
|
| 131 |
|
| 132 | test-repo-xml() {
|
| 133 | # OK these parse
|
| 134 | time find . -name '_chroot' -a -prune -o -name '*.xml' -a -print \
|
| 135 | | $REPO_ROOT/$0 ht8-tool well-formed
|
| 136 | }
|
| 137 |
|
| 138 | # OK we have to skip the <script> tag! And <style>
|
| 139 | #
|
| 140 | # document.location = '#' + params.join('&');
|
| 141 | # gUrlHash = new UrlHash(location.hash);
|
| 142 | #
|
| 143 | # I think textarea we don't though?
|
| 144 |
|
| 145 |
|
| 146 | task-five "$@"
|
| 147 | exit
|
| 148 |
|
| 149 |
|
| 150 | echo '
|
| 151 | In HTML5, instead of
|
| 152 | <script>
|
| 153 | <![CDATA[
|
| 154 | if (x < y) { ... }
|
| 155 | ]]>
|
| 156 | </script>
|
| 157 |
|
| 158 | You can write
|
| 159 |
|
| 160 | <script>
|
| 161 | if (x < y) { ... }
|
| 162 | </script>
|
| 163 |
|
| 164 | <script> <style> <textarea>
|
| 165 |
|
| 166 | These have special escaping rules. I guess we just do NOT lex them at all?
|
| 167 | We can totally SKIP them.
|
| 168 |
|
| 169 | CDATA vs. RCDATA
|
| 170 |
|
| 171 | <textarea>
|
| 172 | <p> <!-- This will show as: <p> -->
|
| 173 | & <!-- This will show as: & -->
|
| 174 | </textarea>
|
| 175 |
|
| 176 | <script>
|
| 177 | <p> <!-- This will show literally as: <p> -->
|
| 178 | & <!-- This will show literally as: & -->
|
| 179 | </script>
|
| 180 |
|
| 181 | The main practical difference is that RCDATA processes HTML entities while
|
| 182 | CDATA treats them as literal text. Both modes ignore HTML tags (treating them
|
| 183 | as plain text) except for their own closing tag. '
|
| 184 | '
|
| 185 |
|