Commit: 2c4d4f2583be270bfc4e4652264f99ec50fb2202 - sfjplib (git)

sfjplib for python

Commit MetaInfo

• Revisão: 2c4d4f2583be270bfc4e4652264f99ec50fb2202 (tree)
• Hora: 2011-07-28 20:00:14
• Autor: Hiromichi MATSUSHIMA <hirom@offi...>
• Commiter: Hiromichi MATSUSHIMA

Mensagem de Log

rewrite form_retriver with htmltree

Mudança Sumário

• modified: form_retriver.py
• modified: test_form_retriver.py

Diff

--- a/form_retriver.py
+++ b/form_retriver.py
@@ -4,126 +4,83 @@
 
 import HTMLParser
 import re
+import htmltree
 
-class Form(dict):
-    def __init__(self, action, method):
-        self.action = action
-        self.method = method
-
-class FormItem(object):
-    def __init__(self, name, attrs, value):
-        self.name = name
-        self.attrs = attrs
-        self.value = value
-
-class FormRetriver(HTMLParser.HTMLParser):
-    UNCLOSE = ("input")
+class Form(list):
+    def __init__(self):
+        self.elements = []
+        self.action = None
+        self.target = None
+        self.enctype = None
+        self.method = None
 
+class FormRetriver(object):
     def __init__(self):
         "Constructor"
-        HTMLParser.HTMLParser.__init__(self)
-        self._current_form = None
-        self._stack = ""
-        self._capture = False
-        self._current_element = None
-        self.forms = []
+        self._forms = []
 
     def parse(self, data):
-        self.feed(data)
-
-    # Handlers
-    def handle_starttag(self, tag, attrs):
-        if tag == "form":
-            attr = dict(attrs)
-            action = attr.get("action", "")
-            method = attr.get("method", "")
-            f = Form(action, method)
-            self._current_form = f
-        elif tag  == "input":
-            self.handle_startendtag(tag, attrs)
-        elif tag == "textarea":
-            self._stack = ""
-            self._capture = True
-            e = FormItem(tag, attrs, "")
-            self._current_element = e
-        elif tag == "option":
-            attr = dict(attrs)
-            val = attr.get("value", "")
-            e = FormItem(tag, attrs, val)
-            if "selected" in attr:
-                self._current_element.value = val
-            if "SELECTED" in attr:
-                self._current_element.value = val
-        elif tag == "select":
-            e = FormItem(tag, attrs, "")
-            self._current_element = e
-            
-    def handle_endtag(self, tag):
-        if tag == "textarea":
-            text = self._stack
-            self._stack = ""
-            self._capture = False
-            e = self._current_element
-            attr = dict(e.attrs)
-            if "NAME" in attr:
-                name = attr["NAME"]
-            elif "name" in attr:
-                name = attr["name"]
-            else:
-                name = ""
-            if name:
-                e.value = text
-                self._current_form[name] = e
-        elif tag == "form":
-            self.forms.append(self._current_form)
-        elif tag == "select":
-            e = self._current_element
-            attr = dict(e.attrs)
-            if "NAME" in attr:
-                name = attr["NAME"]
-            elif "name" in attr:
-                name = attr["name"]
-            else:
-                name = ""
-            if name:
-                self._current_form[name] = e
-
-    def handle_data(self, data):
-        if self._capture:
-            self._stack = self._stack + data
+        tree = htmltree.parse(data)
+        r = tree.root()
+        forms = r.get_elements_by_name("form")
+        for f in forms:
+            self._forms.append(self.form_parse(f))
 
-    def handle_startendtag(self, tag, attrs):
-        if tag == "input":
-            attr = dict(attrs)
-            e = FormItem(tag, attrs, "")
-            if "NAME" in attr:
-                name = attr["NAME"]
-            elif "name" in attr:
-                name = attr["name"]
-            else:
-                name = ""
-            if "VALUE" in attr:
-                e.value = attr["VALUE"]
-            if "value" in attr:
-                e.value = attr["value"]
-
-            if name:
-                self._current_form[name] = e
+    def convert_ref(self, text):
+        f = lambda x: self._convert_ref(x)
+        return re.sub(r"&(\w+);", f, text)
 
-    def handle_entityref(self, name):
-        if self._capture:
-            self._stack = self._stack + self.convert_ref(name)
-
-    def convert_ref(self, name):
+    def _convert_ref(self, m):
         conv_dict = dict(
             lt="<",
             gt=">",
             amp="&",
-            quot="'"
+            quot="'",
+            nbsp=" ",
             )
-        return conv_dict[name]
+        return conv_dict[m.group(1)]
+
+    def form_parse(self, elem):
+        f = Form()
+        f.action = elem.attr("action")
+        f.target = elem.attr("target")
+        f.enctype = elem.attr("enctype")
+        f.method = elem.attr("method")
+        self._r_form_parse(elem, f)
+        return f
+
+    def _r_form_parse(self, elem, f):
+        for e in elem:
+            if e.name == "input":
+                if e.attr("name") == None:
+                    continue
+                f.append((e.attr("name"), e.attr("value")))
+                f.elements.append(e)
+            elif e.name == "textarea":
+                if e.attr("name") == None:
+                    continue
+                t = e.inner_html().encode("utf-8")
+                t = self.convert_ref(t)
+                f.append((e.attr("name"), t))
+                f.elements.append(e)
+            elif e.name == "select":
+                if e.attr("name") == None:
+                    continue
+                name = e.attr("name")
+                for opt in e:
+                    if opt.has_attribute("selected"):
+                        f.append((name, opt.attr("value")))
+                        f.elements.append(opt)
+            elif e.name == "button":
+                if e.attr("name") == None:
+                    continue
+                f.append((e.attr("name"), e.attr("value")))
+                f.elements.append(e)
+            else:
+                self._r_form_parse(e, f)
+        
 
-    def handle_charref(self, ref):
-        if self._capture:
-            self._stack = self._stack + "&#" + ref + ";"
+    def forms(self):
+        return self._forms
         
+    

--- a/test_form_retriver.py
+++ b/test_form_retriver.py
@@ -11,10 +11,10 @@ if __name__ == "__main__":
 
     f = form_retriver.FormRetriver()
     f.parse(html)
-    for form in f.forms:
-        print form.action, form.method, ":"
-        for name in form:
-            print name, ":", form[name].value
+    for form in f.forms():
+        print "\n", form.action, ":"
+        for (key, val) in form:
+            print key, val