Use cmark itself to build spec.html.

Removes build dependency on pandoc.

Closes #256.

Note:  we have lost "smart punctuation," but we can either (a) add
an option to do this in the cmark renderer, or (b) insert unicode
punctuation in the spec as needed.  Not an urgent issue in any case.

1 files changed, 24 insertions, 5 deletions

diff --git a/makespec.py b/makespec.py
index c7b90b5..07bf453 100644
--- a/makespec.py
+++ b/makespec.py
@@ -2,6 +2,7 @@
 import re
 import sys
 from subprocess import *
+from string import Template
 
 if len(sys.argv) == 3:
     specfile = sys.argv[1]
@@ -16,6 +17,18 @@ else:
 def toIdentifier(s):
    return re.sub(r'\s+', '-', re.sub(r'\W+', ' ', s.strip().lower()))
 
+def parseYaml(yaml):
+    metadata = {}
+    def parseField(match):
+        key = match.group(1)
+        val = match.group(2).strip()
+        if re.match(r'^\'', val):
+            val = val[1:len(val) - 1]
+        metadata[key] = val
+    fieldre = re.compile('^(\w+):(.*)$', re.MULTILINE)
+    re.sub(fieldre, parseField, yaml)
+    return metadata
+
 def pipe_through_prog(prog, text):
     p1 = Popen(prog.split(), stdout=PIPE, stdin=PIPE, stderr=PIPE)
     [result, err] = p1.communicate(input=text.encode('utf-8'))
@@ -83,11 +96,14 @@ with open(specfile, 'r', encoding='utf-8') as spec:
                             lastnum[level - 1] = lastnum[level - 1] + 1
                         number = '.'.join([str(x) for x in lastnum])
                     ident = toIdentifier(section)
+                    ln = re.sub(r' ', ' ' + number + ' ', ln, count=1)
                     sections.append(dict(level=level,
                                          contents=section,
                                          ident=ident,
                                          number=number))
                     refs.append("[{0}]: #{1}".format(section, ident))
+                    ln = re.sub(r'# +', '# <a id="{0}"></a> '.format(ident),
+                                ln, count=1)
                 else:
                     ln = re.sub(r'\[([^]]*)\]\(@([^)]*)\)', replaceAnchor, ln)
             else:
@@ -95,22 +111,25 @@ with open(specfile, 'r', encoding='utf-8') as spec:
             mdlines.append(ln)
 
 mdtext = ''.join(mdlines) + '\n\n' + '\n'.join(refs) + '\n'
+yaml = ''.join(yamllines)
+metadata = parseYaml(yaml)
 
 if specformat == "markdown":
-    sys.stdout.write(mdtext)
+    sys.stdout.write(yaml + '\n\n' + mdtext)
 elif specformat == "html":
+    with open("template.html", "r", encoding="utf-8") as templatefile:
+        template = Template(templatefile.read())
     toclines = []
     for section in sections:
         indent = '    ' * (section['level'] - 1)
         toclines.append(indent + '* [' + section['number'] + ' ' +
                         section['contents'] + '](#' + section['ident'] + ')')
     toc = '<div id="TOC">\n\n' + '\n'.join(toclines) + '\n\n</div>\n\n'
-    yaml = ''.join(yamllines) + '\n'
-    prog = "pandoc -s -S --no-highlight --number-sections --template template.html"
-    [retcode, result, err] = pipe_through_prog(prog, yaml + toc + mdtext)
+    prog = "build/src/cmark"
+    [retcode, result, err] = pipe_through_prog(prog, toc + mdtext)
     if retcode == 0:
         result = re.sub(r'␣', '<span class="space"> </span>', result)
-        sys.stdout.write(result)
+        sys.stdout.write(template.substitute(metadata, body=result))
     else:
         sys.stderr.write("Error converting markdown version of spec:\n")
         sys.stderr.write(err)