| 1 | #!/usr/bin/env bash
|
| 2 | #
|
| 3 | # Usage:
|
| 4 | # data_lang/htm8-test.sh
|
| 5 |
|
| 6 | REPO_ROOT=$(cd "$(dirname $0)/.."; pwd)
|
| 7 |
|
| 8 | # Special case: we need $REPO_ROOT
|
| 9 | : ${LIB_OSH=$REPO_ROOT/stdlib/osh}
|
| 10 | source $LIB_OSH/bash-strict.sh
|
| 11 | source $LIB_OSH/task-five.sh
|
| 12 |
|
| 13 | # parse with lazylex/html.py, or data_lang/htm8.py
|
| 14 |
|
| 15 | site-files() {
|
| 16 | #find ../../oilshell/oilshell.org__deploy -name '*.html'
|
| 17 |
|
| 18 | # omit all the _ files
|
| 19 | git ls-files | grep '\.html$'
|
| 20 | }
|
| 21 |
|
| 22 | # Issues with lazylex/html.py
|
| 23 | #
|
| 24 | # - Token ID is annoying to express in Python
|
| 25 | # - re.DOTALL for newlines
|
| 26 | # - can we change that with [.\n]*?
|
| 27 | # - nongreedy match for --> and ?>
|
| 28 |
|
| 29 | ht8-tool() {
|
| 30 | PYTHONPATH="$REPO_ROOT:$REPO_ROOT/vendor" \
|
| 31 | $REPO_ROOT/lazylex/html.py "$@"
|
| 32 | }
|
| 33 |
|
| 34 | test-well-formed() {
|
| 35 | cat >_tmp/bad.html <<EOF
|
| 36 | hi && bye
|
| 37 | EOF
|
| 38 | echo '_tmp/bad.html' | ht8-tool well-formed
|
| 39 | }
|
| 40 |
|
| 41 | # site errors
|
| 42 | #
|
| 43 | # Error in 'release/0.7.pre5/doc/osh-quick-ref.html': (LexError '&& or ||</h4>\n<!-- 2')
|
| 44 | # Error in 'src/data/symbol.html': (LexError "&& mangle[0]!='E' &&")
|
| 45 | # 5833374 tokens in 4710 files
|
| 46 | #
|
| 47 | # The second is the "Woboq" browser, which has CDATA
|
| 48 | # Ah I wonder if we need that.
|
| 49 |
|
| 50 | # Takes ~13 seconds
|
| 51 | test-site() {
|
| 52 | # TODO:
|
| 53 | # - test that the top level lexes
|
| 54 | # - test that each tag lexers
|
| 55 | # - test that each quoted attribute lexes
|
| 56 | # - test that tags are balanced
|
| 57 |
|
| 58 | pushd ../../oilshell/oilshell.org__deploy
|
| 59 |
|
| 60 | # Too many files
|
| 61 | # site-files | xargs wc -l | grep total
|
| 62 |
|
| 63 | # Not using xargs
|
| 64 | time site-files | $REPO_ROOT/$0 ht8-tool well-formed
|
| 65 |
|
| 66 | popd
|
| 67 | }
|
| 68 |
|
| 69 | test-wwz() {
|
| 70 | echo 'TODO: download .wwz from CI'
|
| 71 | }
|
| 72 |
|
| 73 | task-five "$@"
|
| 74 | exit
|
| 75 |
|
| 76 |
|
| 77 | echo '
|
| 78 | In HTML5, instead of
|
| 79 | <script>
|
| 80 | <![CDATA[
|
| 81 | if (x < y) { ... }
|
| 82 | ]]>
|
| 83 | </script>
|
| 84 |
|
| 85 | You can write
|
| 86 |
|
| 87 | <script>
|
| 88 | if (x < y) { ... }
|
| 89 | </script>
|
| 90 |
|
| 91 | <script> <style> <textarea>
|
| 92 |
|
| 93 | These have special escaping rules. I guess we just do NOT lex them at all?
|
| 94 | We can totally SKIP them.
|
| 95 |
|
| 96 | CDATA vs. RCDATA
|
| 97 |
|
| 98 | <textarea>
|
| 99 | <p> <!-- This will show as: <p> -->
|
| 100 | & <!-- This will show as: & -->
|
| 101 | </textarea>
|
| 102 |
|
| 103 | <script>
|
| 104 | <p> <!-- This will show literally as: <p> -->
|
| 105 | & <!-- This will show literally as: & -->
|
| 106 | </script>
|
| 107 |
|
| 108 | The main practical difference is that RCDATA processes HTML entities while
|
| 109 | CDATA treats them as literal text. Both modes ignore HTML tags (treating them
|
| 110 | as plain text) except for their own closing tag. '
|
| 111 | '
|
| 112 |
|