Show More
@@ -0,0 +1,234 b'' | |||||
|
1 | """ | |||
|
2 | IPython extension: %lookfor command for searching docstrings | |||
|
3 | ||||
|
4 | """ | |||
|
5 | # Pauli Virtanen <pav@iki.fi>, 2008. | |||
|
6 | ||||
|
7 | import re, inspect, pkgutil, pydoc | |||
|
8 | ||||
|
9 | #------------------------------------------------------------------------------ | |||
|
10 | # Lookfor functionality | |||
|
11 | #------------------------------------------------------------------------------ | |||
|
12 | ||||
|
13 | # Cache for lookfor: {id(module): {name: (docstring, kind, index), ...}...} | |||
|
14 | # where kind: "func", "class", "module", "object" | |||
|
15 | # and index: index in breadth-first namespace traversal | |||
|
16 | _lookfor_caches = {} | |||
|
17 | ||||
|
18 | # regexp whose match indicates that the string may contain a function signature | |||
|
19 | _function_signature_re = re.compile(r"[a-z_]+\(.*[,=].*\)", re.I) | |||
|
20 | ||||
|
21 | def lookfor(what, modules=None, import_modules=True, regenerate=False): | |||
|
22 | """ | |||
|
23 | Search for objects whose documentation contains all given words. | |||
|
24 | Shows a summary of matching objects, sorted roughly by relevance. | |||
|
25 | ||||
|
26 | Parameters | |||
|
27 | ---------- | |||
|
28 | what : str | |||
|
29 | String containing words to look for. | |||
|
30 | ||||
|
31 | module : str, module | |||
|
32 | Module whose docstrings to go through. | |||
|
33 | import_modules : bool | |||
|
34 | Whether to import sub-modules in packages. | |||
|
35 | Will import only modules in __all__ | |||
|
36 | regenerate: bool | |||
|
37 | Re-generate the docstring cache | |||
|
38 | ||||
|
39 | """ | |||
|
40 | # Cache | |||
|
41 | cache = {} | |||
|
42 | for module in modules: | |||
|
43 | try: | |||
|
44 | c = _lookfor_generate_cache(module, import_modules, regenerate) | |||
|
45 | cache.update(c) | |||
|
46 | except ImportError: | |||
|
47 | pass | |||
|
48 | ||||
|
49 | # Search | |||
|
50 | # XXX: maybe using a real stemming search engine would be better? | |||
|
51 | found = [] | |||
|
52 | whats = str(what).lower().split() | |||
|
53 | if not whats: return | |||
|
54 | ||||
|
55 | for name, (docstring, kind, index) in cache.iteritems(): | |||
|
56 | if kind in ('module', 'object'): | |||
|
57 | # don't show modules or objects | |||
|
58 | continue | |||
|
59 | ok = True | |||
|
60 | doc = docstring.lower() | |||
|
61 | for w in whats: | |||
|
62 | if w not in doc: | |||
|
63 | ok = False | |||
|
64 | break | |||
|
65 | if ok: | |||
|
66 | found.append(name) | |||
|
67 | ||||
|
68 | # Relevance sort | |||
|
69 | # XXX: this is full Harrison-Stetson heuristics now, | |||
|
70 | # XXX: it probably could be improved | |||
|
71 | ||||
|
72 | kind_relevance = {'func': 1000, 'class': 1000, | |||
|
73 | 'module': -1000, 'object': -1000} | |||
|
74 | ||||
|
75 | def relevance(name, docstr, kind, index): | |||
|
76 | r = 0 | |||
|
77 | # do the keywords occur within the start of the docstring? | |||
|
78 | first_doc = "\n".join(docstr.lower().strip().split("\n")[:3]) | |||
|
79 | r += sum([200 for w in whats if w in first_doc]) | |||
|
80 | # do the keywords occur in the function name? | |||
|
81 | r += sum([30 for w in whats if w in name]) | |||
|
82 | # is the full name long? | |||
|
83 | r += -len(name) * 5 | |||
|
84 | # is the object of bad type? | |||
|
85 | r += kind_relevance.get(kind, -1000) | |||
|
86 | # is the object deep in namespace hierarchy? | |||
|
87 | r += -name.count('.') * 10 | |||
|
88 | r += max(-index / 100, -100) | |||
|
89 | return r | |||
|
90 | ||||
|
91 | def relevance_sort(a, b): | |||
|
92 | dr = relevance(b, *cache[b]) - relevance(a, *cache[a]) | |||
|
93 | if dr != 0: return dr | |||
|
94 | else: return cmp(a, b) | |||
|
95 | found.sort(relevance_sort) | |||
|
96 | ||||
|
97 | # Pretty-print | |||
|
98 | s = "Search results for '%s'" % (' '.join(whats)) | |||
|
99 | help_text = [s, "-"*len(s)] | |||
|
100 | for name in found: | |||
|
101 | doc, kind, ix = cache[name] | |||
|
102 | ||||
|
103 | doclines = [line.strip() for line in doc.strip().split("\n") | |||
|
104 | if line.strip()] | |||
|
105 | ||||
|
106 | # find a suitable short description | |||
|
107 | try: | |||
|
108 | first_doc = doclines[0].strip() | |||
|
109 | if _function_signature_re.search(first_doc): | |||
|
110 | first_doc = doclines[1].strip() | |||
|
111 | except IndexError: | |||
|
112 | first_doc = "" | |||
|
113 | help_text.append("%s\n %s" % (name, first_doc)) | |||
|
114 | ||||
|
115 | # Output | |||
|
116 | if len(help_text) > 10: | |||
|
117 | pager = pydoc.getpager() | |||
|
118 | pager("\n".join(help_text)) | |||
|
119 | else: | |||
|
120 | print "\n".join(help_text) | |||
|
121 | ||||
|
122 | def _lookfor_generate_cache(module, import_modules, regenerate): | |||
|
123 | """ | |||
|
124 | Generate docstring cache for given module. | |||
|
125 | ||||
|
126 | Parameters | |||
|
127 | ---------- | |||
|
128 | module : str, None, module | |||
|
129 | Module for which to generate docstring cache | |||
|
130 | import_modules : bool | |||
|
131 | Whether to import sub-modules in packages. | |||
|
132 | Will import only modules in __all__ | |||
|
133 | regenerate: bool | |||
|
134 | Re-generate the docstring cache | |||
|
135 | ||||
|
136 | Returns | |||
|
137 | ------- | |||
|
138 | cache : dict {obj_full_name: (docstring, kind, index), ...} | |||
|
139 | Docstring cache for the module, either cached one (regenerate=False) | |||
|
140 | or newly generated. | |||
|
141 | ||||
|
142 | """ | |||
|
143 | global _lookfor_caches | |||
|
144 | ||||
|
145 | if module is None: | |||
|
146 | module = "numpy" | |||
|
147 | ||||
|
148 | if isinstance(module, str): | |||
|
149 | module = __import__(module) | |||
|
150 | ||||
|
151 | if id(module) in _lookfor_caches and not regenerate: | |||
|
152 | return _lookfor_caches[id(module)] | |||
|
153 | ||||
|
154 | # walk items and collect docstrings | |||
|
155 | cache = {} | |||
|
156 | _lookfor_caches[id(module)] = cache | |||
|
157 | seen = {} | |||
|
158 | index = 0 | |||
|
159 | stack = [(module.__name__, module)] | |||
|
160 | while stack: | |||
|
161 | name, item = stack.pop(0) | |||
|
162 | if id(item) in seen: continue | |||
|
163 | seen[id(item)] = True | |||
|
164 | ||||
|
165 | index += 1 | |||
|
166 | kind = "object" | |||
|
167 | ||||
|
168 | if inspect.ismodule(item): | |||
|
169 | kind = "module" | |||
|
170 | try: | |||
|
171 | _all = item.__all__ | |||
|
172 | except AttributeError: | |||
|
173 | _all = None | |||
|
174 | # import sub-packages | |||
|
175 | if import_modules and hasattr(item, '__path__'): | |||
|
176 | for m in pkgutil.iter_modules(item.__path__): | |||
|
177 | if _all is not None and m[1] not in _all: | |||
|
178 | continue | |||
|
179 | try: | |||
|
180 | __import__("%s.%s" % (name, m[1])) | |||
|
181 | except ImportError: | |||
|
182 | continue | |||
|
183 | for n, v in inspect.getmembers(item): | |||
|
184 | if _all is not None and n not in _all: | |||
|
185 | continue | |||
|
186 | stack.append(("%s.%s" % (name, n), v)) | |||
|
187 | elif inspect.isclass(item): | |||
|
188 | kind = "class" | |||
|
189 | for n, v in inspect.getmembers(item): | |||
|
190 | stack.append(("%s.%s" % (name, n), v)) | |||
|
191 | elif callable(item): | |||
|
192 | kind = "func" | |||
|
193 | ||||
|
194 | doc = inspect.getdoc(item) | |||
|
195 | if doc is not None: | |||
|
196 | cache[name] = (doc, kind, index) | |||
|
197 | ||||
|
198 | return cache | |||
|
199 | ||||
|
200 | #------------------------------------------------------------------------------ | |||
|
201 | # IPython connectivity | |||
|
202 | #------------------------------------------------------------------------------ | |||
|
203 | ||||
|
204 | import IPython.ipapi | |||
|
205 | ip = IPython.ipapi.get() | |||
|
206 | ||||
|
207 | _lookfor_modules = ['numpy', 'scipy'] | |||
|
208 | ||||
|
209 | def lookfor_f(self, arg=''): | |||
|
210 | r""" | |||
|
211 | Search for objects whose documentation contains all given words. | |||
|
212 | Shows a summary of matching objects, sorted roughly by relevance. | |||
|
213 | ||||
|
214 | Usage | |||
|
215 | ----- | |||
|
216 | %lookfor +numpy some words | |||
|
217 | Search module 'numpy' | |||
|
218 | ||||
|
219 | %lookfor_modules numpy scipy | |||
|
220 | Set default modules whose docstrings to search | |||
|
221 | ||||
|
222 | """ | |||
|
223 | lookfor(arg, modules=_lookfor_modules) | |||
|
224 | ||||
|
225 | def lookfor_modules_f(self, arg=''): | |||
|
226 | global _lookfor_modules | |||
|
227 | if not arg: | |||
|
228 | print "Modules included in %lookfor search:", _lookfor_modules | |||
|
229 | else: | |||
|
230 | _lookfor_modules = arg.split() | |||
|
231 | ||||
|
232 | ip.expose_magic('lookfor', lookfor_f) | |||
|
233 | ip.expose_magic('lookfor_modules', lookfor_modules_f) | |||
|
234 |
@@ -1,3 +1,8 b'' | |||||
|
1 | 2008-04-20 Ville Vainio <vivainio@gmail.com> | |||
|
2 | ||||
|
3 | * Extensions/ipy_lookfor.py: add %lookfor magic command | |||
|
4 | (search docstrings for words) by Pauli Virtanen. Close #245. | |||
|
5 | ||||
1 | 2008-04-18 Fernando Perez <Fernando.Perez@berkeley.edu> |
|
6 | 2008-04-18 Fernando Perez <Fernando.Perez@berkeley.edu> | |
2 |
|
7 | |||
3 | * IPython/genutils.py (page): apply workaround to curses bug that |
|
8 | * IPython/genutils.py (page): apply workaround to curses bug that |
General Comments 0
You need to be logged in to leave comments.
Login now