githubmirror/d2l-zh

sphinx_plugin.py
222 строки8 KB
muli
add reg back and modified
14 сен 2017, 06:42
14 сен 2017, 06:42768c545
"""
1. Converting markdown files into jupyter notebooks
2. Remove filename headers, such as from P01-C01-xx.ipynb to xx.ipynb
"""
import notedown
import glob
import pkg_resources
import nbformat
import re
import shutil
import os
import time
import tarfile
from zipfile import ZipFile

# timeout in second to evaluate a notebook
timeout = 240
# limit the number of lines in a cell output
max_output_length = 500
# the files will be ingored for execution
ignore_execution = []

def _replace_ext(fname, new_ext):
    """replace the file extension in a filename"""
    parts = fname.split('.')
    if len(parts) <= 1:
        return fname
    parts[-1] = new_ext
    return '.'.join(parts)

def _get_new_fname(fname):
    """chapter01-something/haha.ipynb -> haha.ipynb"""
    header_re = re.compile("(chapter[\d\-\w]+/)(.*)")
    m = header_re.match(fname)
    return m.groups()[1] if m else fname

def _has_output(notebook):
    """if a notebook contains output"""
    for cell in notebook.cells:
        if 'outputs' in cell and cell['outputs']:
            return True
    return False

def convert_md():
    """Find all markdown files, convert into jupyter notebooks
    """
    converted_files = []
    reader = notedown.MarkdownReader(match='strict')
    files = glob.glob('*/*.md')
    # evaluate the newest file first, so we can catchup error ealier
    files.sort(key=os.path.getmtime, reverse=True)

    do_eval = int(os.environ.get('DO_EVAL', True))
    if do_eval:
        do_eval = int(os.environ.get('EVAL', True))

    if not do_eval:
        print('=== Will skip evaluating notebooks')

    for fname in files:
        new_fname = _get_new_fname(fname)
        # parse if each markdown file is actually a jupyter notebook
        with open(fname, 'r') as fp:
            valid = '```{.python .input' in fp.read()
            if not valid:
                if new_fname != fname:
                    print('=== Rename %s -> %s' % (fname, new_fname))
                    shutil.copyfile(fname, new_fname)
                    converted_files.append((fname, new_fname))
                continue

        # read
        with open(fname, 'r') as f:
            notebook = reader.read(f)

        if do_eval and not (_has_output(notebook) or
                any([i in fname for i in ignore_execution])):
            print('=== Evaluate %s with timeout %d sec'%(fname, timeout))
            tic = time.time()
            # update from ../data to data
            for c in notebook.cells:
                if c.get('cell_type', None) == 'code':
                    c['source'] = c['source'].replace(
                        '"../data', '"data').replace("'../data", "'data")
            notedown.run(notebook, timeout)
            print('=== Finished in %f sec'%(time.time()-tic))

        # even that we will check it later, but do it ealier so we can see the
        # error message before evaluating all notebooks
        _check_notebook(notebook)

        # write
        # need to add language info to for syntax highlight
        notebook['metadata'].update({'language_info':{'name':'python'}})
        new_fname = _replace_ext(new_fname, 'ipynb')
        print('=== Convert %s -> %s' % (fname, new_fname))
        with open(new_fname, 'w') as f:
            f.write(nbformat.writes(notebook))

        converted_files.append((fname, new_fname))
    return converted_files

def rename_ipynb():
    """renmae all ipynb files"""
    renamed_files = []
    for fname in glob.glob('*.ipynb'):
        new_fname = _get_new_fname(fname)
        if fname != new_fname:
            print('=== Rename %s -> %s' % (fname, new_fname))
            shutil.copyfile(fname, new_fname)
            renamed_files.append((fname, new_fname))
    return renamed_files

def update_links(app, docname, source):
    """Update all C01-P01-haha.md into haha.html"""
    def _new_url(m):
        assert len(m.groups()) == 1, m
        url = m.groups()[0]
        if url.startswith('./'):
            url = url[2:]
        if url.startswith('../'):
            url = url[3:]
        if _get_new_fname(url) != url:
            url = _replace_ext(_get_new_fname(url), 'html')
        return url

    for i,j in enumerate(source):
        if os.path.exists(docname+'.md') or os.path.exists(docname+'.ipynb'):
            source[i] = re.sub('\]\(([\w/.-]*)\)',
                               lambda m : ']('+_new_url(m)+')', j)
        elif os.path.exists(docname+'.rst'):
            source[i] = re.sub('\<([\w/.-]*)\>`\_',
                                   lambda m: '<'+_new_url(m)+'>`_', j)

def _check_notebook(notebook):
    # TODO(mli) lint check
    for cell in notebook.cells:
         if 'outputs' in cell:
             src = cell['source']
             nlines = 0
             try:
                 for o in cell['outputs']:
                     if 'text' in o:
                         nlines += len(o['text'].split('\n'))
                     assert 'traceback' not in o, '%s, %s'%(o['ename'], o['evalue'])
                 assert nlines < max_output_length, 'Too long cell output'
             except AssertionError:
                 print('This cell\'s output contains error:\n')
                 print('-'*40)
                 print(src)
                 print('-'*40)
                 raise

def check_output(app, exception):
    for fname in glob.glob('*.ipynb'):
        print('=== Check '+fname)

        with open(fname, 'r') as f:
            nb = nbformat.read(f, as_version=4)
            _check_notebook(nb)

def _release_notebook(dst_dir):
    """convert .md into notebooks and make a zip file"""
    reader = notedown.MarkdownReader(match='strict')
    files = glob.glob('*/*.md')
    package_files = ['environment.yml', 'utils.py', 'README.md', 'LICENSE']
    package_files.extend(glob.glob('img/*'))
    package_files.extend(glob.glob('data/*'))
    for fname in files:
        # parse if each markdown file is actually a jupyter notebook
        with open(fname, 'r') as fp:
            valid = '```{.python .input' in fp.read()
            if not valid:
                package_files.append(fname)
                continue
        # read
        with open(fname, 'r') as f:
            notebook = reader.read(f)
        # write
        new_fname = _replace_ext(fname, 'ipynb')
        with open(new_fname, 'w') as f:
            f.write(nbformat.writes(notebook))
        package_files.append(new_fname)

    print('=== Packing ', package_files)

    with ZipFile(os.path.join(dst_dir, 'gluon_tutorials_zh.zip'), 'w') as pkg:
        for f in package_files:
            pkg.write(f)

    with tarfile.open(
            os.path.join(dst_dir, 'gluon_tutorials_zh.tar.gz'), "w:gz") as tar:
        for f in package_files:
            tar.add(f)

    for f in glob.glob('*/*.ipynb'):
        os.remove(f)

converted_files = convert_md()
renamed_files = rename_ipynb()
ignore_list = [f for f,_ in converted_files + renamed_files]

def remove_generated_files(app, exception):
    for _, f in renamed_files + converted_files:
        print('=== Remove %s' % (f))
        os.remove(f)

def release_notebook(app, exception):
    _release_notebook(app.builder.outdir)

def generate_htaccess(app, exception):
    print('=== Generate .htaccess file')
    with open(app.builder.outdir + '/.htaccess', 'w') as f:
        f.write('ErrorDocument 404 https://zh.gluon.ai/404.html\n')
        # force to use https
        f.write('RewriteEngine On\n')
        f.write('RewriteCond %{SERVER_PORT} 80\n')
        f.write('RewriteRule ^(.*)$ https://zh.gluon.ai/$1 [R,L]\n')
        for old, new in renamed_files + converted_files:
            f.write('Redirect /%s /%s\n'%(
                _replace_ext(old, 'html'), _replace_ext(new, 'html')
            ))