Implement PyPI screenscraping for EasyInstall. Fix a bug in requirement

version checking.  Document new options for screen scraping.

--HG--
branch : setuptools
extra : convert_revision : svn%3A6015fed2-1504-0410-9fe1-9d1591cc4771/sandbox/trunk/setuptools%4041041

4 files changed, 164 insertions, 125 deletions

diff --git a/EasyInstall.txt b/EasyInstall.txt
index 6635320e..2a1d875b 100755
--- a/EasyInstall.txt
+++ b/EasyInstall.txt
@@ -210,10 +210,42 @@ Command-Line Options
     URL or filename, so that the installer will not be confused by the presence
     of multiple ``setup.py`` files in the build directory.
 
+``--scan-url=URL, -s URL`` (New in 0.4a1)
+    Scan the specified "download page" for direct links to downloadable eggs or
+    source distributions.  Any usable packages will be downloaded if they are
+    required by a command line argument.  For example, this::
+
+        easy_install -s http://peak.telecommunity.com/dist PyProtocols
+
+    will download and install the latest version of PyProtocols linked from
+    the PEAK downloads page, but ignore the other download links on that page.
+
+    You may use this option more than once, to list multiple download pages.
+    If all requested packages can be found using the specified download pages,
+    the Python Package Index will *not* be consulted.
+
+``--index-url=URL, -u URL`` (New in 0.4a1)
+    Specifies the base URL of the Python Package Index.  The default is
+    http://www.python.org/pypi if not specified.  When a package is requested
+    that is not locally available or linked from a ``--scan-url`` download
+    page, the package index will be searched for download pages for the needed
+    package, and those download pages will be searched for links to download
+    an egg or source distribution.
+
 
 Release Notes/Change History
 ============================
 
+0.4a1
+ * Added ``--scan-url`` and ``--index-url`` options, to scan download pages
+   and search PyPI for needed packages.
+
+ * Fixed a bug in requirements processing for exact versions (i.e. ``==`` and
+   ``!=``) when only one condition was included.
+
+ * Added ``safe_name()`` and ``safe_version()`` APIs to clean up handling of
+   arbitrary distribution names and versions found on PyPI.
+
 0.3a4
  * ``pkg_resources`` now supports resource directories, not just the resources
    in them.  In particular, there are ``resource_listdir()`` and
diff --git a/easy_install.py b/easy_install.py
index 00e316ff..81415677 100755
--- a/easy_install.py
+++ b/easy_install.py
@@ -28,98 +28,133 @@ import __builtin__
 from distutils.sysconfig import get_python_lib
 from shutil import rmtree   # must have, because it can be called from __del__
 from pkg_resources import *
-
 _os = sys.modules[os.name]
 _open = open
 
-
-
-
-
-
-
-
-EXTENSIONS = (
-   (EGG_DIST,    ".egg"),
-   (SOURCE_DIST, ".tar.gz"),
-   (SOURCE_DIST, ".tar.bz2"),
-   (SOURCE_DIST, ".tar"),
-   (SOURCE_DIST, ".zip"),
-   (SOURCE_DIST, ".tgz"),
-)
-
-class URLDistribution(Distribution):
-    """A distribution that has not been installed"""
-
-    def __init__(self, url, metadata=None):
-        path = urlparse.urlparse(url)[2]
-        base = path.split('/')[-1]
-
-        for typecode, ext in EXTENSIONS:
-            if base.endswith(ext):
-                base = base[:-len(ext)]
-                break
-        else:
-            raise DistributionNotFound(url)
-
-        self.typecode = typecode
-        name, version, py_version, platform = [None]*4
-        match = pkg_resources.EGG_NAME(base)
-        if match:
-            name,version,py_version,platform = match.group(
-                'name','ver','pyver','plat'
-            )
-        else:
-            name = base
-        Distribution.__init__(self,
-            url, metadata=metadata, name=name, version=version or "0",
-            py_version=py_version or pkg_resources.PY_MAJOR, platform=platform
+class Opener(urllib.FancyURLopener):
+    def http_error_default(self, url, fp, errcode, errmsg, headers):
+        """Default error handling -- don't raise an exception."""
+        info = urllib.addinfourl(fp, headers, "http:" + url)
+        info.status, info.reason = errcode, errmsg
+        return info
+opener = Opener()
+
+HREF = re.compile(r"""href\s*=\s*['"]?([^'"> ]+)""", re.I)
+EXTENSIONS = ".tar.gz .tar.bz2 .tar .zip .tgz".split()
+
+def distros_for_url(url, metadata=None):
+    """Yield egg or source distribution objects that might be found at a URL"""
+
+    path = urlparse.urlparse(url)[2]
+    base = urllib.unquote(path.split('/')[-1])
+
+    if base.endswith('.egg'):
+        dist = Distribution.from_filename(base, metadata)
+        dist.path = url
+        yield dist
+        return  # only one, unambiguous interpretation
+
+    for ext in EXTENSIONS:
+        if base.endswith(ext):
+            base = base[:-len(ext)]
+            break
+    else:
+        return  # no extension matched
+
+    # Generate alternative interpretations of a source distro name
+    # Because some packages are ambiguous as to name/versions split
+    # e.g. "adns-python-1.1.0", "egenix-mx-commercial", etc.
+    # So, we generate each possible interepretation (e.g. "adns, python-1.1.0"
+    # "adns-python, 1.1.0", and "adns-python-1.1.0, no version").  In practice,
+    # the spurious interpretations should be ignored, because in the event
+    # there's also an "adns" package, the spurious "python-1.1.0" version will
+    # compare lower than any numeric version number, and is therefore unlikely
+    # to match a request for it.  It's still a potential problem, though, and
+    # in the long run PyPI and the distutils should go for "safe" names and
+    # versions in source distribution names.
+
+    parts = base.split('-')
+    for p in range(1,len(parts)+1):
+        yield Distribution(
+            url, metadata, '-'.join(parts[:p]), '-'.join(parts[p:]),
+            distro_type = SOURCE_DIST
         )
 
-
-
-
-
 class PackageIndex(AvailableDistributions):
     """A distribution index that scans web pages for download URLs"""
 
     def __init__(self,index_url="http://www.python.org/pypi",*args,**kw):
         AvailableDistributions.__init__(self,*args,**kw)
-        self.index_url = index_url
+        self.index_url = index_url + "/"[:not index_url.endswith('/')]
         self.scanned_urls = {}
+        self.fetched_urls = {}
+        self.package_pages = {}
 
     def scan_url(self, url):
         self.process_url(url, True)
 
     def process_url(self, url, retrieve=False):
-        if url in self.scanned_urls:
-            return
-        try:
-            dist = URLDistribution(url)
-        except DistributionNotFound:    # not a distro, so scan the page
-            if not retrieve:
-                return    # unless we're skipping retrieval
-        else:
-            # It's a distro, just process it
-            self.scanned_urls[url] = True
-            self.add(dist)  # XXX should check py_ver/platform!
+        if url in self.scanned_urls and not retrieve:
             return
 
-        f = urllib.urlopen(url)
         self.scanned_urls[url] = True
+        dists = list(distros_for_url(url))
+        map(self.add, dists)
+
+        if dists or not retrieve or url in self.fetched_urls:
+            # don't need the actual page
+            return
 
+        f = opener.open(url)
+        self.fetched_urls[url] = self.fetched_urls[f.url] = True
         if 'html' not in f.headers['content-type'].lower():
             f.close()   # not html, we can't process it
             return
-        url = f.url     # handle redirects
-        href = re.compile(r"""href\s*=\s*['"]?([^'"> ]+)""", re.I)
+
+        base = f.url     # handle redirects
         page = f.read()
         f.close()
-        for match in href.finditer(page):
-            link = urlparse.urljoin(url, match.group(1))
-            self.process_url(link)
-
-
+        if url.startswith(self.index_url):
+            self.process_index(url, page)
+        else:
+            for match in HREF.finditer(page):
+                link = urlparse.urljoin(base, match.group(1))
+                self.process_url(link)
+
+    def find_packages(self,requirement):       
+        self.scan_url(self.index_url + requirement.distname)
+        if not self.package_pages.get(requirement.key):
+            # We couldn't find the target package, so search the index page too
+            self.scan_url(self.index_url)
+        for url in self.package_pages.get(requirement.key,()):
+            # scan each page that might be related to the desired package
+            self.scan_url(url)
+
+    def process_index(self,url,page):
+        def scan(link):
+            if link.startswith(self.index_url):
+                parts = map(
+                    urllib.unquote, link[len(self.index_url):].split('/')
+                )
+                if len(parts)==2:
+                    # it's a package page, sanitize and index it
+                    pkg = safe_name(parts[0])
+                    ver = safe_version(parts[1])
+                    self.package_pages.setdefault(pkg.lower(),{})[link] = True          
+        if url==self.index_url or 'Index of Packages</title>' in page:
+            # process an index page into the package-page index
+            for match in HREF.finditer(page):
+                scan( urlparse.urljoin(url, match.group(1)) )
+        else:
+            scan(url)   # ensure this page is in the page index
+            # process individual package page
+            for tag in ("<th>Home Page", "<th>Download URL"):
+                pos = page.find(tag)
+                if pos!=-1:
+                    match = HREF.search(page,pos)
+                    if match:
+                        # Process the found URL
+                        self.scan_url(urlparse.urljoin(url, match.group(1)))
 
     def obtain(self,requirement):
         self.find_packages(requirement)
@@ -127,41 +162,6 @@ class PackageIndex(AvailableDistributions):
             if dist in requirement:
                 return dist
 
-    def find_packages(self,requirement):
-        pass # XXX process PyPI entries for package
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
 class Installer:
     """Manage a download/build/install process"""
 
@@ -801,11 +801,11 @@ def main(argv, factory=Installer):
 
     parser.add_option("-u", "--index-url", dest="index_url", metavar="URL",
                       default="http://www.python.org/pypi",
-                      help="Base URL of Python Package Index")
+                      help="base URL of Python Package Index")
 
     parser.add_option("-s", "--scan-url", dest="scan_urls", metavar="URL",
                       action="append",
-                      help="Additional URL(s) to search for packages")
+                      help="additional URL(s) to search for packages")
 
     (options, args) = parser.parse_args()
 
diff --git a/pkg_resources.py b/pkg_resources.py
index 927b7942..a9697bbc 100644
--- a/pkg_resources.py
+++ b/pkg_resources.py
@@ -54,6 +54,9 @@ class InvalidOption(ResolutionError):
 _provider_factories = {}
 PY_MAJOR = sys.version[:3]
 
+EGG_DIST    = 2
+SOURCE_DIST = 1
+
 def register_loader_type(loader_type, provider_factory):
     """Register `provider_factory` to make providers for `loader_type`
 
@@ -77,9 +80,6 @@ def get_provider(moduleName):
 
 
 
-
-
-
 def get_platform():
     """Return this platform's string for platform-specific distributions
 
@@ -1146,23 +1146,21 @@ def parse_version(s):
 
 
 
-EGG_DIST    = 2
-SOURCE_DIST = 1
-
 class Distribution(object):
     """Wrap an actual or potential sys.path entry w/metadata"""
-    typecode = EGG_DIST
+    
     def __init__(self,
         path_str, metadata=None, name=None, version=None,
-        py_version=PY_MAJOR, platform=None
+        py_version=PY_MAJOR, platform=None, distro_type = EGG_DIST
     ):
         if name:
-            self.name = name.replace('_','-')
+            self.name = safe_name(name)
         if version:
-            self._version = version.replace('_','-')
+            self._version = safe_version(version)
         self.py_version = py_version
         self.platform = platform
         self.path = path_str
+        self.distro_type = distro_type
         self.metadata = metadata
 
     def installed_on(self,path=None):
@@ -1187,6 +1185,8 @@ class Distribution(object):
         )
     from_filename = classmethod(from_filename)
 
+
+
     # These properties have to be lazy so that we don't have to load any
     # metadata until/unless it's actually needed.  (i.e., some distributions
     # may not know their name or version without loading PKG-INFO)
@@ -1330,7 +1330,7 @@ def parse_requirements(strs):
 
 
 def _sort_dists(dists):
-    tmp = [(dist.version,dist.typecode,dist) for dist in dists]
+    tmp = [(dist.version,dist.distro_type,dist) for dist in dists]
     tmp.sort()
     dists[::-1] = [d for v,t,d in tmp]
 
@@ -1382,16 +1382,16 @@ class Requirement:
             item = item.parsed_version
         elif isinstance(item,basestring):
             item = parse_version(item)
-        last = True
+        last = None
         for parsed,trans,op,ver in self.index:
             action = trans[cmp(item,parsed)]
             if action=='F':     return False
             elif action=='T':   return True
             elif action=='+':   last = True
-            elif action=='-':   last = False
+            elif action=='-' or last is None:   last = False
+        if last is None: last = True    # no rules encountered
         return last
 
-
     def __hash__(self):
         return self.__hash
 
@@ -1414,7 +1414,7 @@ state_machine = {
     '>' :  'F+F',
     '>=':  'T+F',
     '==':  'T..',
-    '!=':  'F..',
+    '!=':  'F++',
 }
 
 
diff --git a/setuptools/tests/test_resources.py b/setuptools/tests/test_resources.py
index 88b5f889..91199f47 100644
--- a/setuptools/tests/test_resources.py
+++ b/setuptools/tests/test_resources.py
@@ -258,12 +258,19 @@ class RequirementsTests(TestCase):
                             ImmutableSet(["foo","bar"])))
         )
 
-
-
-
-
-
-
+    def testVersionEquality(self):
+        r1 = Requirement.parse("setuptools==0.3a2")
+        r2 = Requirement.parse("setuptools!=0.3a4")
+        d = Distribution.from_filename
+
+        self.failIf(d("setuptools-0.3a4.egg") in r1)
+        self.failIf(d("setuptools-0.3a1.egg") in r1)
+        self.failIf(d("setuptools-0.3a4.egg") in r2)
+
+        self.failUnless(d("setuptools-0.3a2.egg") in r1)
+        self.failUnless(d("setuptools-0.3a2.egg") in r2)
+        self.failUnless(d("setuptools-0.3a3.egg") in r2)
+        self.failUnless(d("setuptools-0.3a5.egg") in r2)