#!python # ***** BEGIN LICENSE BLOCK ***** # Version: MPL 1.1/GPL 2.0/LGPL 2.1 # # The contents of this file are subject to the Mozilla Public License # Version 1.1 (the "License"); you may not use this file except in # compliance with the License. You may obtain a copy of the License at # http://www.mozilla.org/MPL/ # # Software distributed under the License is distributed on an "AS IS" # basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See the # License for the specific language governing rights and limitations # under the License. # # The Original Code is Komodo code. # # The Initial Developer of the Original Code is ActiveState Software Inc. # Portions created by ActiveState Software Inc are Copyright (C) 2000-2007 # ActiveState Software Inc. All Rights Reserved. # # Contributor(s): # ActiveState Software Inc # # Alternatively, the contents of this file may be used under the terms of # either the GNU General Public License Version 2 or later (the "GPL"), or # the GNU Lesser General Public License Version 2.1 or later (the "LGPL"), # in which case the provisions of the GPL or the LGPL are applicable instead # of those above. If you wish to allow use of your version of this file only # under the terms of either the GPL or the LGPL, and not to allow others to # use your version of this file under the terms of the MPL, indicate your # decision by deleting the provisions above and replace them with the notice # and other provisions required by the GPL or the LGPL. If you do not delete # the provisions above, a recipient may use your version of this file under # the terms of any one of the MPL, the GPL or the LGPL. # # ***** END LICENSE BLOCK ***** """The langzone of the codeintel database. See the database/database.py module docstring for an overview. """ import sys import os from os.path import (join, dirname, exists, expanduser, splitext, basename, split, abspath, isabs, isdir, isfile, normpath) import threading import time from glob import glob from pprint import pprint, pformat import logging from cStringIO import StringIO import codecs import copy import apsw # for SQLite symbol database import ciElementTree as ET from codeintel2.common import * from codeintel2 import util from codeintel2.database.util import rmdir from codeintel2.database.langlibbase import LangDirsLibBase #---- globals log = logging.getLogger("codeintel.db") #log.setLevel(logging.DEBUG) #---- Database zone and lib implementations class LangDirsLib(LangDirsLibBase): """A zone providing a view into an ordered list of dirs in a db/$lang/... area of the db. These are dished out via Database.get_lang_lib(), which indirectly then is dished out by the LangZone.get_lib(). Mostly this is just a view on the LangZone singleton for this particular language. Dev Notes: - The goal is to provide quick has_blob() and get_blob() -- i.e. some caching is involved (if 'foo' referred to 'some/path/to/foo.py' a minute ago then it still does). As well, scanning/loading is done automatically as necessary. For example, if a request for Perl blob 'Bar' is made but there is no 'Bar' in the database yet, this code looks for a 'Bar.pm' on the file system and will scan it, load it and return the blob for it. """ def __init__(self, lang_zone, lock, lang, name, dirs): LangDirsLibBase.__init__(self) self.lang_zone = lang_zone self._lock = lock self.mgr = lang_zone.mgr self.lang = lang self.name = name self.dirs = dirs self.import_handler \ = self.mgr.citadel.import_handler_from_lang(self.lang) self._blob_imports_from_prefix_cache = {} self._importables_from_dir_cache = {} # We keep a "weak" merged cache of blobname lookup for all dirs # in this zone -- where "weak" means that we verify a hit by # checking the current real blob_index for that dir (which may # have changed). This caching slows down lookup for single-dir # LangDirsZones, but should scale better for LangDirsZones with # many dirs. (TODO-PERF: test this assertion.) self._dir_and_blobbase_from_blobname = {} def __repr__(self): return "<%s %s>" % (self.lang, self.name) def _acquire_lock(self): self._lock.acquire() def _release_lock(self): self._lock.release() def has_blob(self, blobname, ctlr=None): dbsubpath = self._dbsubpath_from_blobname(blobname, ctlr=ctlr) return dbsubpath is not None def has_blob_in_db(self, blobname, ctlr=None): """Return true if the blobname is in the database. Typically this method is only used for debugging and .has_blob() is what you want. """ dbsubpath = self._dbsubpath_from_blobname( blobname, ctlr=ctlr, only_look_in_db=True) return dbsubpath is not None def get_blob(self, blobname, ctlr=None): self._acquire_lock() try: dbsubpath = self._dbsubpath_from_blobname(blobname, ctlr=ctlr) if dbsubpath is not None: return self.lang_zone.load_blob(dbsubpath) else: return None finally: self._release_lock() def get_blob_imports(self, prefix): """Return the set of imports under the given prefix. "prefix" is a tuple of import name parts. E.g. ("xml", "sax") for "import xml.sax." in Python. Or ("XML", "Parser") for "use XML::Parser::" in Perl. See description in database.py docstring for details. """ self._acquire_lock() try: if prefix not in self._blob_imports_from_prefix_cache: if prefix: for dir in self.dirs: importables = self._importables_from_dir(dir) if prefix[0] in importables: sub_importables = self._importables_from_dir( join(dir, *prefix)) imports = set( (name, is_dir_import) for name, (_, _, is_dir_import) in sub_importables.items() ) break else: imports = set() else: imports = set() for dir in self.dirs: importables = self._importables_from_dir(dir) imports.update( (name, is_dir_import) for name, (_, _, is_dir_import) in importables.items() ) self._blob_imports_from_prefix_cache[prefix] = imports return self._blob_imports_from_prefix_cache[prefix] finally: self._release_lock() def blobs_with_basename(self, basename, ctlr=None): """Return all blobs that match the given base path. I.e. a filename lookup across all files in the dirs of this lib. "basename" is a string, e.g. 'Http.js' "ctlr" (optional) is an EvalController instance. If specified it should be used in the normal way (logging, checking .is_aborted()). A "blob" is a global scope-tag hit in all of the blobs for the execution set buffers. Returns the empty list if no hits. """ self.ensure_all_dirs_scanned(ctlr=ctlr) blobs = [] # we can't use self.get_blob because that only returns one answer; we # we need all of them. self._acquire_lock() try: for dir in self.dirs: dbfile_from_blobname = self.lang_zone.dfb_from_dir(dir, {}) blobbase = dbfile_from_blobname.get(basename) if blobbase is not None: dhash = self.lang_zone.dhash_from_dir(dir) dbsubpath = join(dhash, blobbase) blobs.append(self.lang_zone.load_blob(dbsubpath)) finally: self._release_lock() return blobs def hits_from_lpath(self, lpath, ctlr=None, curr_buf=None): """Return all hits of the given lookup path. I.e. a symbol table lookup across all files in the dirs of this lib. "lpath" is a lookup name list, e.g. ['Casper', 'Logging'] or ['dojo', 'animation']. "ctlr" (optional) is an EvalController instance. If specified it should be used in the normal way (logging, checking .is_aborted()). "curr_buf" (optional), if specified, is the current buf for which this query is being made. Hits from it should be skipped (i.e. don't bother searching it). A "hit" is (, ). Each one represent a scope-tag or variable-tag hit in all of the blobs for the execution set buffers. Returns the empty list if no hits. """ assert isinstance(lpath, tuple) # common mistake to pass in a string # Need to have (at least once) scanned all importables. # Responsibility for ensuring the scan data is *up-to-date* # is elsewhere. self.ensure_all_dirs_scanned(ctlr=ctlr) if curr_buf: curr_blobname = curr_buf.blob_from_lang.get(self.lang, {}).get("name") curr_buf_dir = dirname(curr_buf.path) # Naive implementation (no caching) hits = [] for dir in self.dirs: if ctlr and ctlr.is_aborted(): log.debug("ctlr aborted") break toplevelname_index = self.lang_zone.load_index( dir, "toplevelname_index", {}) for blobname in toplevelname_index.get_blobnames(lpath[0], ()): if curr_buf and curr_buf_dir == dir and blobname == curr_blobname: continue blob = self.get_blob(blobname, ctlr=ctlr) try: elem = blob for p in lpath: #LIMITATION: *Imported* names at each scope are # not being included here. This is fine while we # just care about JavaScript. if curr_buf: if "__file_local__" in elem.get("attributes", "").split(): # this is a file-local element in a different blob, # don't look at it raise KeyError elem = elem.names[p] except KeyError: continue hits.append( (elem, (blob, list(lpath[:-1]))) ) return hits def toplevel_cplns(self, prefix=None, ilk=None, ctlr=None): """Return completion info for all top-level names matching the given prefix and ilk in all blobs in this lib. "prefix" is a 3-character prefix with which to filter top-level names. If None (or not specified), results are not filtered based on the prefix. "ilk" is a symbol type (e.g. "class", "variable", "function") with which to filter results. If None (or not specified), results of any ilk are returned. "ctlr" (optional) is an EvalController instance. If specified it should be used in the normal way (logging, checking .is_aborted()). Returns a list of 2-tuples: (, ). Note: the list is not sorted, because often some special sorting is required for the different completion evaluators that might use this API. """ self.ensure_all_dirs_scanned(ctlr=ctlr) cplns = [] # Naive implementation (no caching) for dir in self.dirs: if ctlr and ctlr.is_aborted(): log.debug("ctlr aborted") break try: toplevelname_index = self.lang_zone.load_index( dir, "toplevelname_index") except EnvironmentError: # No toplevelname_index for this dir likely indicates that # there weren't any files of the current lang in this dir. continue cplns += toplevelname_index.toplevel_cplns(prefix=prefix, ilk=ilk) return cplns def _importables_from_dir(self, dir): if dir not in self._importables_from_dir_cache: self._importables_from_dir_cache[dir] \ = self.import_handler.find_importables_in_dir(dir) return self._importables_from_dir_cache[dir] def _dbsubpath_from_blobname(self, blobname, ctlr=None, only_look_in_db=False): """Return the subpath to the dbfile for the given blobname, or None if not found. Remember that this is complicated by possible multi-level imports. E.g. "import foo.bar" or "import foo" where 'foo' refers to 'foo/__init__.py'. """ assert blobname is not None, "'blobname' cannot be None" lang_zone = self.lang_zone self._acquire_lock() try: # Use our weak cache to try to return quickly. if blobname in self._dir_and_blobbase_from_blobname: blobdir, blobbase \ = self._dir_and_blobbase_from_blobname[blobname] # Check it. The actual info for that dir may have changed. dbfile_from_blobname = lang_zone.dfb_from_dir(blobdir) if blobbase in dbfile_from_blobname: log.debug("have blob '%s' in '%s'? yes (in weak cache)", blobname, blobdir) return join(lang_zone.dhash_from_dir(blobdir), dbfile_from_blobname[blobbase]) # Drop from weak cache. del self._dir_and_blobbase_from_blobname[blobname] # Brute force: look in each dir. blobparts = blobname.split(self.import_handler.sep) blobbase = blobparts[-1] for dir in self.dirs: if ctlr and ctlr.is_aborted(): log.debug("aborting search for blob '%s' on %s: " "ctlr aborted", blobname, self) return None # Is the blob in 'blobdir' (i.e. a non-multi-level import # that has been scanned already). blobdir = join(dir, *blobparts[:-1]) dbfile_from_blobname = lang_zone.dfb_from_dir(blobdir, {}) if self.lang == "Perl": # Perl uses the full blob name - not just the blob base, # see bug 89106 for details. if blobname in dbfile_from_blobname: self._dir_and_blobbase_from_blobname[blobname] \ = (blobdir, blobname) log.debug("have blob '%s' in '%s'? yes (in dir index)", blobname, blobdir) return join(lang_zone.dhash_from_dir(blobdir), dbfile_from_blobname[blobname]) if blobbase in dbfile_from_blobname: self._dir_and_blobbase_from_blobname[blobname] \ = (blobdir, blobbase) log.debug("have blob '%s' in '%s'? yes (in dir index)", blobname, blobdir) return join(lang_zone.dhash_from_dir(blobdir), dbfile_from_blobname[blobbase]) importables = self._importables_from_dir(blobdir) # 'importables' look like, for Python: # {'foo': ('foo.py', None, False), # 'pkg': ('pkg/__init__.py', '__init__', False)} # for Perl: # {'LWP': ('LWP.pm', None, True), # 'File': (None, None, True)} # | | | `-- is-dir-import # | | `-- subdir-blobbase # | `-- blobfile # `-- blobbase if blobbase not in importables: continue blobfile, subdir_blobbase, is_dir_import = importables[blobbase] if blobfile is None: # There isn't an actual importable file here -- just # a dir prefix to a multidir import. log.debug("have blob '%s' in %s? no", blobname, self) continue elif os.sep in blobfile: # This is an import from a subdir. We need to get a new dbf. blobdir = join(blobdir, dirname(blobfile)) blobfile = basename(blobfile) blobbase = subdir_blobbase dbfile_from_blobname = lang_zone.dfb_from_dir(blobdir, {}) if blobbase in dbfile_from_blobname: self._dir_and_blobbase_from_blobname[blobname] \ = (blobdir, blobbase) log.debug("have blob '%s' in '%s'? yes (in dir index)", blobname, blobdir) return join(lang_zone.dhash_from_dir(blobdir), dbfile_from_blobname[blobbase]) # The file isn't loaded. if not only_look_in_db: log.debug("%s importables in '%s':\n %s", self.lang, blobdir, importables) log.debug("'%s' likely provided by '%s' in '%s': " "attempting load", blobname, blobfile, blobdir) try: buf = self.mgr.buf_from_path( join(blobdir, blobfile), self.lang) except (EnvironmentError, CodeIntelError), ex: # This can occur if the path does not exist, such as a # broken symlink, or we don't have permission to read # the file, or the file does not contain text. continue buf.scan_if_necessary() dbfile_from_blobname = lang_zone.dfb_from_dir(blobdir, {}) if self.lang == "Perl": # Perl uses the full blob name - not just the blob base, # see bug 89106 for details. if blobname in dbfile_from_blobname: self._dir_and_blobbase_from_blobname[blobname] \ = (blobdir, blobname) log.debug("have blob '%s' in '%s'? yes (in dir index)", blobname, blobdir) return join(lang_zone.dhash_from_dir(blobdir), dbfile_from_blobname[blobname]) if blobbase in dbfile_from_blobname: self._dir_and_blobbase_from_blobname[blobname] \ = (blobdir, blobbase) log.debug("have blob '%s' in '%s'? yes (after load)", blobname, blobdir) return join(lang_zone.dhash_from_dir(blobdir), dbfile_from_blobname[blobbase]) log.debug("have blob '%s' in %s? no", blobname, self) return None finally: self._release_lock() class LangTopLevelNameIndex(object): """A wrapper around the plain-dictionary toplevelname_index for a LangZone dir to provide better performance for continual updating and some simpler access. {ilk -> toplevelname -> blobnames} # Problem A 'toplevelname_index' is a merge of {blobname -> ilk -> toplevelnames} data for all resources in its dir. As those resources are continually re-scanned (e.g. as a file is edited in Komodo), it would be too expensive to update this index everytime. # Solution Keep a list of "recent updates" and only merge them into the main data when that buf hasn't been updated in "a while" and when needed for saving the index. Note: Buffer *removals* are not put on-deck, but removed immediately. # .get_blobnames(..., ilk=None) Originally the toplevelname_index stored {toplevelname -> blobnames}. The per-"ilk" level was added afterwards to support occassional ilk filtering for PHP (and possible eventually other langs). .get_blobnames() still behaves like a {toplevelname -> blobnames} mapping, but it provides an optional "ilk" keyword arg to limit the results to that ilk. # Notes on locking This class does not guard its datastructures with locking. It is up to the LangZone using this to guard against simultaneous access on separate threads. """ def __init__(self, data=None, timeout=90): # toplevelname_index data: {ilk -> toplevelname -> blobnames} if data is None: self._data = {} else: self._data = data # Time (in seconds) to hold a change "on deck". # Timed-out changes are merged on .get() and .update(). self.timeout = timeout self._on_deck = { # basename # the basename of the buf path # -> [timestamp, # time of the last update # # The dict in res_index, a.k.a. 'res_data' # {blobname -> ilk -> toplevelnames}, # # Lazily generated pivot, a.k.a. 'res_data_pivot' # {ilk -> toplevelname -> blobnames} # ] } def __repr__(self): num_toplevelnames = sum(len(v) for v in self._data.itervalues()) return ("" % (num_toplevelnames, len(self._on_deck))) def merge(self): """Merge all on-deck changes with `self.data'.""" for base, (timestamp, res_data, res_data_pivot) in self._on_deck.items(): if res_data_pivot is None: res_data_pivot = self._pivot_res_data(res_data) # res_data_pivot: {ilk -> toplevelname -> blobnames} # "bft" means blobnames_from_toplevelname for ilk, bft in res_data_pivot.iteritems(): data_bft = self._data.setdefault(ilk, {}) for toplevelname, blobnames in bft.iteritems(): if toplevelname not in data_bft: data_bft[toplevelname] = blobnames else: data_bft[toplevelname].update(blobnames) del self._on_deck[base] def merge_expired(self, now): """Merge expired on-deck changes with `self.data'.""" for base, (timestamp, res_data, res_data_pivot) in self._on_deck.items(): if now - timestamp < self.timeout: continue if res_data_pivot is None: res_data_pivot = self._pivot_res_data(res_data) # res_data_pivot: {ilk -> toplevelname -> blobnames} # "bft" means blobnames_from_toplevelname for ilk, bft in res_data_pivot.iteritems(): data_bft = self._data.setdefault(ilk, {}) for toplevelname, blobnames in bft.iteritems(): if toplevelname not in data_bft: data_bft[toplevelname] = blobnames else: data_bft[toplevelname].update(blobnames) del self._on_deck[base] @property def data(self): self.merge() return self._data def update(self, base, old_res_data, new_res_data): now = time.time() self.remove(base, old_res_data) self._on_deck[base] = [now, new_res_data, None] self.merge_expired(now) def remove(self, base, old_res_data): if base in self._on_deck: del self._on_deck[base] else: # Remove old refs from current data. # old_res_data: {blobname -> ilk -> toplevelnames} # self._data: {ilk -> toplevelname -> blobnames} for blobname, toplevelnames_from_ilk in old_res_data.iteritems(): for ilk, toplevelnames in toplevelnames_from_ilk.iteritems(): for toplevelname in toplevelnames: try: self._data[ilk][toplevelname].remove(blobname) except KeyError: pass # ignore this for now, might indicate corruption else: if not self._data[ilk][toplevelname]: del self._data[ilk][toplevelname] if not self._data.get(ilk): del self._data[ilk] def _pivot_res_data(self, res_data): # res_data: {blobname -> ilk -> toplevelnames} # res_data_pivot: {ilk -> toplevelname -> blobnames} res_data_pivot = {} for blobname, toplevelnames_from_ilk in res_data.iteritems(): for ilk, toplevelnames in toplevelnames_from_ilk.iteritems(): pivot_bft = res_data_pivot.setdefault(ilk, {}) for toplevelname in toplevelnames: if toplevelname not in pivot_bft: pivot_bft[toplevelname] = set([blobname]) else: pivot_bft[toplevelname].add(blobname) return res_data_pivot def toplevel_cplns(self, prefix=None, ilk=None): """Return completion info for all top-level names matching the given prefix and ilk. "prefix" is a 3-character prefix with which to filter top-level names. If None (or not specified), results are not filtered based on the prefix. "ilk" is a symbol type (e.g. "class", "variable", "function") with which to filter results. If None (or not specified), results of any ilk are returned. Returns a list of 2-tuples: (, ). """ self.merge_expired(time.time()) # Need to check merged and on-deck items: cplns = [] # ...on-deck items for base, (timestamp, res_data, res_data_pivot) in self._on_deck.items(): if res_data_pivot is None: res_data_pivot = self._on_deck[base][2] \ = self._pivot_res_data(res_data) # res_data_pivot: {ilk -> toplevelname -> blobnames} if ilk is None: for i, bft in res_data_pivot.iteritems(): cplns += [(i, toplevelname) for toplevelname in bft] elif ilk in res_data_pivot: cplns += [(ilk, toplevelname) for toplevelname in res_data_pivot[ilk]] # ...merged data # self._data: {ilk -> toplevelname -> blobnames} if ilk is None: for i, bft in self._data.iteritems(): cplns += [(i, toplevelname) for toplevelname in bft] elif ilk in self._data: cplns += [(ilk, toplevelname) for toplevelname in self._data[ilk]] # Naive implementation: Instead of maintaining a separate # 'toplevelprefix_index' (as we do for StdLibsZone and CatalogsZone) # for now we'll just gather all results and filter on the prefix # here. Only if this proves to be a perf issue will we add the # complexity of another index: # {ilk -> prefix -> toplevelnames} if prefix is not None: cplns = [(i, t) for i, t in cplns if t.startswith(prefix)] return cplns def get_blobnames(self, toplevelname, default=None, ilk=None): """Return the blobnames defining the given toplevelname. If "ilk" is given then only symbols of that ilk will be considered. If not match is found the "default" is returned. """ self.merge_expired(time.time()) blobnames = set() # First check on-deck items. for base, (timestamp, res_data, res_data_pivot) in self._on_deck.items(): if res_data_pivot is None: res_data_pivot = self._on_deck[base][2] \ = self._pivot_res_data(res_data) # res_data_pivot: {ilk -> toplevelname -> blobnames} if ilk is None: for bft in res_data_pivot.itervalues(): if toplevelname in bft: blobnames.update(bft[toplevelname]) elif ilk in res_data_pivot: if toplevelname in res_data_pivot[ilk]: blobnames.update(res_data_pivot[ilk][toplevelname]) #TODO: Put lookup in merged data ahead of lookup in on-deck -- so # we don't do on-deck work if not necessary. # Then, fallback to already merged data. # self._data: {ilk -> toplevelname -> blobnames} if ilk is None: for bft in self._data.itervalues(): if toplevelname in bft: blobnames.update(bft[toplevelname]) elif ilk in self._data: if toplevelname in self._data[ilk]: blobnames.update(self._data[ilk][toplevelname]) if blobnames: return blobnames return default class LangZone(object): """Singleton zone managing a particular db/$lang/... area. # caching and memory control We cache all retrieved indices and blobs and maintain their latest access time. To try to manage memory consumption, we rely on a bookkeeper thread (the indexer) to periodically call .cull_mem() -- which unloads cache items that have not been accessed in a while. (TODO: - Get the indexer to actually call .cull_mem() and .save() periodically. - Test that .cull_mem() actually results in the process releasing memory.) # robustness (TODO) Work should be done to improve robustness. - Collect filesystem interactions in one place. - Rationalize OSError handling. - Consider a journal system, if necessary/feasible. My hope is to get away without one and rely on graceful recovery. The db does not store critical info so can allow some loss of data (it can all be regenerated). """ toplevelname_index_class = LangTopLevelNameIndex def __init__(self, mgr, lang): self.mgr = mgr self.db = mgr.db self.lang = lang self.base_dir = join(self.db.base_dir, "db", util.safe_lang_from_lang(lang)) self._check_lang(lang) self._hook_handlers = self.mgr.hook_handlers_from_lang(lang) self._lock = threading.RLock() self._dhash_from_dir_cache = {} self._dirslib_cache = {} # We cache the set of recent indeces and blobs in memory. # {db-subpath: [index-object, ]), # ...} # For example: # {'7bce640bc48751b128af5c8bf5df8412/res_index': # [, 1158289000]), # ...} self._index_and_atime_from_dbsubpath = {} #TODO-PERF: Use set() object for this? Compare perf. self._is_index_dirty_from_dbsubpath = {} # set of dirty indeces ##TODO: blob caching and *use* this #self._blob_and_atime_from_dbsubpath = {} #XXX Need a 'dirty-set' for blobs? No, because currently # .update_buf_data() saves blob changes to disk immediately. Not # sure that is best for perf. Definitely not ideal for the # "editset". def __repr__(self): return "<%s lang db>" % self.lang def _acquire_lock(self): self._lock.acquire() def _release_lock(self): self._lock.release() def _check_lang(self, lang): """Ensure that the given lang matches case exactly with the lang in the db. If this invariant is broken, then weird things with caching can result. """ if exists(self.base_dir): lang_path = join(self.base_dir, "lang") try: fin = open(lang_path, 'r') except EnvironmentError, ex: self.db.corruption("LangZone._check_lang", "could not open `%s': %s" % (lang_path, ex), "recover") fin = open(lang_path, 'w') try: fin.write(lang) finally: fin.close() else: try: lang_on_disk = fin.read().strip() finally: fin.close() assert lang_on_disk == lang #TODO: If Database.dhash_from_dir() grows caching, then this # shouldn't bother. def dhash_from_dir(self, dir): if dir not in self._dhash_from_dir_cache: self._dhash_from_dir_cache[dir] = self.db.dhash_from_dir(dir) return self._dhash_from_dir_cache[dir] def dfb_from_dir(self, dir, default=None): """Get the {blobname -> dbfile} mapping index for the given dir. 'dfb' stands for 'dbfile_from_blobname'. This must be called with the lock held. """ return self.load_index(dir, "blob_index", default) def get_buf_scan_time(self, buf): #TODO Canonicalize path (or assert that it is canonicalized) self._acquire_lock() try: dir, base = split(buf.path) res_index = self.load_index(dir, "res_index", {}) if base not in res_index: return None return res_index[base][0] finally: self._release_lock() def get_buf_data(self, buf): #TODO Canonicalize path (or assert that it is canonicalized) # Should have a Resource object that we pass around that # handles all of this. self._acquire_lock() try: dir, base = split(buf.path) res_index = self.load_index(dir, "res_index", {}) if base not in res_index: raise NotFoundInDatabase("%s buffer '%s' not found in database" % (buf.lang, buf.path)) scan_time, scan_error, res_data = res_index[base] blob_from_lang = {} if res_data: try: dbfile_from_blobname = self.dfb_from_dir(dir) except EnvironmentError, ex: # DB corruption will be noted in remove_buf_data() self.remove_buf_data(buf) raise NotFoundInDatabase("%s buffer '%s' not found in database" % (buf.lang, buf.path)) dhash = self.dhash_from_dir(dir) for blobname in res_data: dbsubpath = join(dhash, dbfile_from_blobname[blobname]) try: blob = self.load_blob(dbsubpath) except ET.XMLParserError, ex: self.db.corruption("LangZone.get_buf_data", "could not parse dbfile for '%s' blob: %s"\ % (blobname, ex), "recover") self.remove_buf_data(buf) raise NotFoundInDatabase( "`%s' buffer `%s' blob was corrupted in database" % (buf.path, blobname)) except EnvironmentError, ex: self.db.corruption("LangZone.get_buf_data", "could not read dbfile for '%s' blob: %s"\ % (blobname, ex), "recover") self.remove_buf_data(buf) raise NotFoundInDatabase( "`%s' buffer `%s' blob not found in database" % (buf.path, blobname)) lang = blob.get("lang") assert lang is not None blob_from_lang[lang] = blob return scan_time, scan_error, blob_from_lang finally: self._release_lock() def remove_path(self, path): """Remove the given resource from the database.""" #TODO Canonicalize path (or assert that it is canonicalized) # Should have a Resource object that we pass around that # handles all of this. self._acquire_lock() try: dir, base = split(path) res_index = self.load_index(dir, "res_index", {}) try: scan_time, scan_error, res_data = res_index[base] except KeyError: # This resource isn't loaded in the db. Nothing to remove. return try: blob_index = self.load_index(dir, "blob_index") except EnvironmentError, ex: self.db.corruption("LangZone.remove_path", "could not read blob_index for '%s' dir: %s" % (dir, ex), "recover") blob_index = {} is_hits_from_lpath_lang = self.lang in self.db.import_everything_langs if is_hits_from_lpath_lang: try: toplevelname_index = self.load_index(dir, "toplevelname_index") except EnvironmentError, ex: self.db.corruption("LangZone.remove_path", "could not read toplevelname_index for '%s' dir: %s" % (dir, ex), "recover") toplevelname_index = self.toplevelname_index_class() dhash = self.dhash_from_dir(dir) del res_index[base] for blobname in res_data: try: dbfile = blob_index[blobname] except KeyError: blob_index_path = join(dhash, "blob_index") self.db.corruption("LangZone.remove_path", "'%s' blob not in '%s'" \ % (blobname, blob_index_path), "ignore") continue del blob_index[blobname] for path in glob(join(self.base_dir, dhash, dbfile+".*")): log.debug("fs-write: remove %s blob file '%s/%s'", self.lang, dhash, basename(path)) os.remove(path) if is_hits_from_lpath_lang: toplevelname_index.remove(base, res_data) self.changed_index(dir, "res_index") self.changed_index(dir, "blob_index") if is_hits_from_lpath_lang: self.changed_index(dir, "toplevelname_index") finally: self._release_lock() #TODO Database.clean() should remove dirs that have no # blob_index entries. def remove_buf_data(self, buf): """Remove the given buffer from the database.""" self.remove_path(buf.path) def _get_symbols_db_conn(self): """ Returns a connection to the SQLite3 database that houses known symbols. If the database does not exist yet, create it. This database is used for quick "Go to Symbol"-type lookups and is located in the root of the "codeintel/" directory. @return apsw.Connection """ try: apsw except NameError: return None symbols_db = join(self.db.base_dir, "symbols.db") exists = isfile(symbols_db) try: # apsw.Connection() requires a utf-8-encoded filename, which is not # always the case -- particularly on Windows. conn = apsw.Connection(symbols_db.decode(sys.getfilesystemencoding())) except: log.exception("Unable to create/open symbols.db") return None cursor = conn.cursor() if not exists: cursor.execute(""" CREATE TABLE symbols ('symbol_id' INTEGER PRIMARY KEY, 'symbol' TEXT NOT NULL, 'type' TEXT NOT NULL, 'filepath' TEXT NOT NULL, 'lineno' INTEGER NOT NULL, 'lang' TEXT NOT NULL, 'parent' TEXT) """) cursor.execute("CREATE UNIQUE INDEX 'symbol_id' on symbols (symbol_id ASC)") try: cursor.execute("COMMIT") except apsw.SQLError: log.warn("Unable to create symbols.db: possibly exists?") return None return conn def _update_symbols_db(self, conn, action, blob): """ Update the symbols in the SQLite3 symbol database contained in the given blob's filename. @param conn The SQLite3 database connection returned by `self._get_symbols_db_conn()`. @param action The database action to perform, either "add", "remove", or "update". @param blob The blob that contains the file whose symbols are to be stored in the database. """ if not conn: return cursor = conn.cursor() cursor.execute("BEGIN") # Delete any symbols in the blob's file. if action == "update" or action == "remove": cursor.execute("DELETE FROM symbols WHERE filepath=?", (blob.get("src"),)) # Add all of blob's symbols recursively. if action == "add" or action == "update": def insert_all(elem, parent=None): """ Inserts the given element (assuming it is a valid symbol) and all of its valid children into the symbol database. @param elem The element to insert. @param parent Optional parent of `elem` for context. """ symbol = elem.get("name", "") type = elem.get("ilk", "") src = blob.get("src", "") lineno = elem.get("line", "") lang = blob.get("lang") valid = symbol and type and src and lineno and \ type != 'argument' if valid: cursor.execute("INSERT INTO symbols VALUES (?, ?, ?, ?, ?, ?, ?)", (None, symbol, type, src, lineno, lang, parent and parent.get("name", "") or None)) for n in elem: insert_all(n, valid and elem) insert_all(blob) try: cursor.execute("commit") except apsw.SQLError: log.error("Unable to save to symbols db.") def _close_symbols_db_conn(self, conn): """ Closes the connection to the SQLite3 symbol database. @param conn The SQLite3 database connection returned by `self._get_symbols_db_conn()`. """ if conn: conn.close() def update_buf_data(self, buf, scan_tree, scan_time, scan_error, skip_scan_time_check=False): """Update this LangZone with the buffer data. @param buf {CitadelBuffer} the buffer whose data is being added to the database. @param scan_tree {ciElementTree} the CIX scan data. Might be None if there was an early scanning failure. @param scan_time {timestamp} the time of the scan, typically the mtime of the file @param scan_error {str} an error string if scanning failed, or None if it was succesful. @param skip_scan_time_check {boolean} (default False) is a boolean indicating if the buffer data should be updated even if `scan_time` is <= that in the database. """ self._acquire_lock() try: #TODO: Canonicalize path (or assert that it is canonicalized) dir, base = split(buf.path) # Get the current data, if any. res_index = self.load_index(dir, "res_index", {}) res_index_has_changed = False blob_index = self.load_index(dir, "blob_index", {}) blob_index_has_changed = False is_hits_from_lpath_lang = self.lang in self.db.import_everything_langs if is_hits_from_lpath_lang: #TODO: Not sure {} for a default is correct here. toplevelname_index = self.load_index(dir, "toplevelname_index", {}) toplevelname_index_has_changed = False try: (old_scan_time, old_scan_error, old_res_data) = res_index[base] except KeyError: # adding a new entry (old_scan_time, old_scan_error, old_res_data) = None, None, {} else: # updating an existing entry if not skip_scan_time_check and scan_time is not None \ and scan_time <= old_scan_time: log.debug("skipping db update for '%s': %s < %s and " "no 'skip_scan_time_check' option", base, scan_time, old_scan_time) return log.debug("update from %s buf '%s'", buf.lang, buf.path) # Parse the tree and get the list of blobnames. # res_data: {blobname -> ilk -> toplevelnames} new_res_data = {} new_blobnames_and_blobs = [] if scan_tree: for blob in scan_tree[0]: lang = blob.get("lang") assert blob.get("lang") == self.lang, "'%s' != '%s' (blob %r)" % (blob.get("lang"), self.lang, blob) blobname = blob.get("name") toplevelnames_from_ilk = new_res_data.setdefault(blobname, {}) for toplevelname, elem in blob.names.iteritems(): if "__file_local__" in elem.get("attributes", "").split(): # don't put file local things in toplevel names continue ilk = elem.get("ilk") or elem.tag if ilk not in toplevelnames_from_ilk: toplevelnames_from_ilk[ilk] = set([toplevelname]) else: toplevelnames_from_ilk[ilk].add(toplevelname) new_blobnames_and_blobs.append((blobname, blob)) # Determine necessary changes to res_index. if scan_error: if (scan_time != old_scan_time or scan_error != old_scan_error): res_index[base] = (scan_time, scan_error, old_res_data) res_index_has_changed = True else: # Only consider new blobs if there wasn't a scan error. # I.e., we want to preserve the last good scan info. if (scan_time != old_scan_time or scan_error != old_scan_error or new_res_data != old_res_data): res_index[base] = (scan_time, scan_error, new_res_data) res_index_has_changed = True if is_hits_from_lpath_lang: if new_res_data != old_res_data: toplevelname_index.update(base, old_res_data, new_res_data) toplevelname_index_has_changed = True # Determine necessary changes to blob_index and the # dbfiles and then make them. dbfile_changes = [] for blobname, blob in new_blobnames_and_blobs: if blobname in old_res_data: dbfile_changes.append(("update", blobname, blob)) else: dbfile_changes.append(("add", blobname, blob)) for blobname in old_res_data: if blobname not in new_res_data: dbfile_changes.append(("remove", blobname, None)) dhash = self.dhash_from_dir(dir) conn = self._get_symbols_db_conn() for action, blobname, blob in dbfile_changes: if action == "add": dbfile = self.db.bhash_from_blob_info( buf.path, self.lang, blobname) blob_index[blobname] = dbfile blob_index_has_changed = True dbdir = join(self.base_dir, dhash) if not exists(dbdir): self._mk_dbdir(dbdir, dir) #XXX What to do on write failure? log.debug("fs-write: %s blob '%s/%s'", self.lang, dhash, dbfile) if blob.get("src") is None: blob.set("src", buf.path) # for defns_from_pos() support ET.ElementTree(blob).write(join(dbdir, dbfile+".blob")) elif action == "remove": dbfile = blob_index[blobname] del blob_index[blobname] blob_index_has_changed = True #XXX What to do on removal failure? log.debug("fs-write: remove %s blob '%s/%s'", self.lang, dhash, dbfile) os.remove(join(self.base_dir, dhash, dbfile+".blob")) elif action == "update": # Try to only change the dbfile on disk if it is # different. s = StringIO() if blob.get("src") is None: blob.set("src", buf.path) # for defns_from_pos() support ET.ElementTree(blob).write(s) new_dbfile_content = s.getvalue() dbfile = blob_index[blobname] dbpath = join(self.base_dir, dhash, dbfile+".blob") # PERF: Might be nice to cache the new dbfile # content for the next time this resource is # updated. For files under edit this will be # common. I.e. just for the "editset". try: fin = open(dbpath, 'r') except (OSError, IOError), ex: # Technically if the dbfile doesn't exist, this # is a sign of database corruption. No matter # though (for this blob anyway), we are about to # replace it. old_dbfile_content = None else: try: old_dbfile_content = fin.read() finally: fin.close() if new_dbfile_content != old_dbfile_content: if not exists(dirname(dbpath)): self._mk_dbdir(dirname(dbpath), dir) #XXX What to do if fail to write out file? log.debug("fs-write: %s blob '%s/%s'", self.lang, dhash, dbfile) fout = open(dbpath, 'w') try: fout.write(new_dbfile_content) finally: fout.close() self._update_symbols_db(conn, action, blob) self._close_symbols_db_conn(conn) if res_index_has_changed: self.changed_index(dir, "res_index") if blob_index_has_changed: self.changed_index(dir, "blob_index") if is_hits_from_lpath_lang and toplevelname_index_has_changed: self.changed_index(dir, "toplevelname_index") finally: self._release_lock() #TODO Database.clean() should remove dirs that have no # blob_index entries. def _mk_zone_skel(self): log.debug("fs-write: mkdir '%s'", self.base_dir) os.makedirs(self.base_dir) log.debug("fs-write: create 'lang'") fout = codecs.open(join(self.base_dir, "lang"), 'wb', 'utf-8') try: fout.write(self.lang) finally: fout.close() def _mk_dbdir(self, dbdir, dir): if not exists(self.base_dir): self._mk_zone_skel() log.debug("fs-write: mkdir '%s'", dbdir[len(self.base_dir)+1:]) os.mkdir(dbdir) log.debug("fs-write: '%s/path'", dbdir[len(self.base_dir)+1:]) fout = codecs.open(join(dbdir, "path"), 'wb', 'utf-8') try: fout.write(dir) finally: fout.close() def load_blob(self, dbsubpath): """This must be called with the lock held.""" log.debug("TODO: LangZone.load_blob: add blob caching!") log.debug("fs-read: load %s blob '%s'", self.lang, dbsubpath) dbpath = join(self.base_dir, dbsubpath+".blob") blob = ET.parse(dbpath).getroot() for hook_handler in self._hook_handlers: try: hook_handler.post_db_load_blob(blob) except: log.exception("error running hook: %r.post_db_load_blob(%r)", hook_handler, blob) return blob def load_index(self, dir, index_name, default=None): """Get the indicated index. "dir" is the dir path this index represents. "index_name" is the name of the index. "default" (default None) indicate the value to return for the index if the index doesn't exist. If not set (or None) then an OSError is raised if the index doesn't exist. The index is loaded from a pickle on disk, if necessary, put into the cache system, and returned. This must be called with the lock held. """ self._acquire_lock() try: dbsubpath = join(self.db.dhash_from_dir(dir), index_name) # If index path is in the cache: return it, update its atime. now = time.time() if dbsubpath in self._index_and_atime_from_dbsubpath: log.debug("cache-read: load %s index '%s'", self.lang, dbsubpath) self._index_and_atime_from_dbsubpath[dbsubpath][1] = now return self._index_and_atime_from_dbsubpath[dbsubpath][0] # Otherwise, load it. log.debug("fs-read: load %s index '%s'", self.lang, dbsubpath) dbpath = join(self.base_dir, dbsubpath) index = self.db.load_pickle(dbpath, default) if index_name == "toplevelname_index": index = self.toplevelname_index_class(index) self._index_and_atime_from_dbsubpath[dbsubpath] = [index, now] return index finally: self._release_lock() def changed_index(self, dir, index_name): """Note that we've changed this index (so it can be saved as appropriate). """ self._acquire_lock() try: now = time.time() dbsubpath = join(self.db.dhash_from_dir(dir), index_name) self._index_and_atime_from_dbsubpath[dbsubpath][1] = now self._is_index_dirty_from_dbsubpath[dbsubpath] = True finally: self._release_lock() def save_index(self, dbsubpath, index): if isinstance(index, self.toplevelname_index_class): index = index.data self.db.save_pickle(join(self.base_dir, dbsubpath), index) def save(self): self._acquire_lock() try: for dbsubpath in self._is_index_dirty_from_dbsubpath: self.save_index(dbsubpath, self._index_and_atime_from_dbsubpath[dbsubpath][0]) self._is_index_dirty_from_dbsubpath = {} finally: self._release_lock() def cull_mem(self): """Drop indeces and tree from cache that have not been accessed in over 5 minutes. To attempt to keep memory consumption under control we want to ensure we don't keep everything cached from the db in memory until process completion. The plan is to have a thread periodically cull memory. """ #TOTEST: Does Python/Komodo actually release this memory or # are we kidding ourselves? log.debug("LangZone: culling memory") TIME_SINCE_ACCESS = 300.0 # 5 minutes since last access self._acquire_lock() try: N = 30 if len(self._index_and_atime_from_dbsubpath) < N: # Too few indeces in memory to bother culling. return now = time.time() for dbsubpath, (index, atime) \ in self._index_and_atime_from_dbsubpath.items(): if now - atime > TIME_SINCE_ACCESS: if dbsubpath in self._is_index_dirty_from_dbsubpath: self.save_index(dbsubpath, index) del self._is_index_dirty_from_dbsubpath[dbsubpath] del self._index_and_atime_from_dbsubpath[dbsubpath] except: log.exception("Exception culling memory") finally: self._release_lock() #XXX Database.clean(): Go through each $lang/dir/res_index and # clean out files in the index but that don't actually exist # anymore. #XXX Database.clean(): drop memory for indeces that are quite # old (say haven't been accessed in 20 minutes). #XXX Database.check(): Shouldn't have too many cached indeces in # memory. How old is the oldest one? Estimate memory size # used by all loaded indeces? # TODO: When a directory no longer exists on the filesystem - should we # 1) remove the db data, or # 2) mark it as expired. # Option 2 would work better for (network) mounted filesystems, as it # could just be an intermittent issue. def clean(self): """Clean out any expired/old codeintel information.""" base_dir = self.base_dir if not exists(base_dir): return for d in os.listdir(base_dir): path_path = join(base_dir, d, "path") if not exists(path_path): continue path = codecs.open(path_path, encoding="utf-8").read() if not exists(path): # Referenced directory no longer exists - so remove the db info. log.debug("clean:: scanned directory no longer exists: %r", path) rmdir(join(base_dir, d)) def get_lib(self, name, dirs): """ Dev Notes: We make a lib for a particular sequence of dirs a singleton because: 1. The sequence of dirs for a language's import path tends to not change, so the same object will tend to get used. 2. This allows caching of filesystem lookups to be done naturally on the LangDirsLib instance. To ensure that this cache doesn't grow unboundedly we only allow there to be N cached LangDirsLib's. A good value for N is when there are relatively few cache misses. Ideally we'd want to count the number of cache misses (i.e. LangDirsLib instance creations) for a number of "typical" uses of codeintel -- i.e. a long running Komodo profile. Failing that we'll just use N=10. """ assert isinstance(dirs, (tuple, list)) canon_dirs = tuple(set(abspath(normpath(expanduser(d))) for d in dirs)) if canon_dirs in self._dirslib_cache: return self._dirslib_cache[canon_dirs] langdirslib = LangDirsLib(self, self._lock, self.lang, name, canon_dirs) # Ensure that these directories are all *up-to-date*. langdirslib.ensure_all_dirs_scanned() self._dirslib_cache[canon_dirs] = langdirslib return langdirslib def reportMemory(self): """ Report on memory usage from this LangZone. @returns {dict} memory usage; keys are the paths, values are a dict of "amount" -> number "units" -> "bytes" | "count" "desc" -> str description """ log.debug("%s LangZone: reporting memory", self.lang) import memutils return { "komodo\\codeintel/langzone/%s/index-cache" % (self.lang,): { "amount": len(self._index_and_atime_from_dbsubpath), "units": "count", "desc": "Number of cached indices.", }, "explicit/python/codeintel/%s/index-cache" % (self.lang,): { "amount": memutils.memusage(self._index_and_atime_from_dbsubpath), "units": "bytes", "desc": "The number of bytes of %s codeintel index caches." % (self.lang,), }, }