|
|
#gq:
|
|
|
# encoding: utf-8
|
|
|
|
|
|
from __future__ import absolute_import, division, print_function, unicode_literals
|
|
|
# 未来兼容导入：保证代码在 Python2/3 下行为一致
|
|
|
|
|
|
import json
|
|
|
import os
|
|
|
import re
|
|
|
import shutil
|
|
|
import threading
|
|
|
import warnings
|
|
|
|
|
|
import six
|
|
|
from django.conf import settings
|
|
|
from django.core.exceptions import ImproperlyConfigured
|
|
|
from datetime import datetime
|
|
|
from django.utils.encoding import force_str
|
|
|
from haystack.backends import BaseEngine, BaseSearchBackend, BaseSearchQuery, EmptyResults, log_query
|
|
|
from haystack.constants import DJANGO_CT, DJANGO_ID, ID
|
|
|
from haystack.exceptions import MissingDependency, SearchBackendError, SkipDocument
|
|
|
from haystack.inputs import Clean, Exact, PythonData, Raw
|
|
|
from haystack.models import SearchResult
|
|
|
from haystack.utils import get_identifier, get_model_ct
|
|
|
from haystack.utils import log as logging
|
|
|
from haystack.utils.app_loading import haystack_get_model
|
|
|
from jieba.analyse import ChineseAnalyzer # 中文分词器
|
|
|
from whoosh import index
|
|
|
from whoosh.analysis import StemmingAnalyzer
|
|
|
from whoosh.fields import BOOLEAN, DATETIME, IDLIST, KEYWORD, NGRAM, NGRAMWORDS, NUMERIC, Schema, TEXT
|
|
|
from whoosh.fields import ID as WHOOSH_ID
|
|
|
from whoosh.filedb.filestore import FileStorage, RamStorage
|
|
|
from whoosh.highlight import ContextFragmenter, HtmlFormatter
|
|
|
from whoosh.highlight import highlight as whoosh_highlight
|
|
|
from whoosh.qparser import QueryParser
|
|
|
from whoosh.searching import ResultsPage
|
|
|
from whoosh.writing import AsyncWriter
|
|
|
|
|
|
try:
|
|
|
import whoosh
|
|
|
except ImportError:
|
|
|
raise MissingDependency(
|
|
|
"The 'whoosh' backend requires the installation of 'Whoosh'. Please refer to the documentation.")
|
|
|
|
|
|
# Handle minimum requirement.
|
|
|
if not hasattr(whoosh, '__version__') or whoosh.__version__ < (2, 5, 0):
|
|
|
raise MissingDependency(
|
|
|
"The 'whoosh' backend requires version 2.5.0 or greater.")
|
|
|
|
|
|
# Bubble up the correct error.
|
|
|
|
|
|
# 用于解析 whoosh 存储的 datetime 字符串的正则
|
|
|
DATETIME_REGEX = re.compile(
|
|
|
'^(?P<year>\d{4})-(?P<month>\d{2})-(?P<day>\d{2})T(?P<hour>\d{2}):(?P<minute>\d{2}):(?P<second>\d{2})(\.\d{3,6}Z?)?$')
|
|
|
|
|
|
# 线程本地存储：用于在内存模式时共享 RamStorage
|
|
|
LOCALS = threading.local()
|
|
|
LOCALS.RAM_STORE = None
|
|
|
|
|
|
|
|
|
class WhooshHtmlFormatter(HtmlFormatter):
|
|
|
"""
|
|
|
自定义高亮 HTML 输出格式，保持与其他后端（Solr、ES 等）一致。
|
|
|
模板：<em>高亮文本</em>
|
|
|
"""
|
|
|
template = '<%(tag)s>%(t)s</%(tag)s>'
|
|
|
|
|
|
|
|
|
class WhooshSearchBackend(BaseSearchBackend):
|
|
|
# Whoosh 保留字，查询时需转义或避免
|
|
|
RESERVED_WORDS = (
|
|
|
'AND',
|
|
|
'NOT',
|
|
|
'OR',
|
|
|
'TO',
|
|
|
)
|
|
|
|
|
|
# Whoosh 保留字符，同样需转义
|
|
|
RESERVED_CHARACTERS = (
|
|
|
'\\', '+', '-', '&&', '||', '!', '(', ')', '{', '}',
|
|
|
'[', ']', '^', '"', '~', '*', '?', ':', '.',
|
|
|
)
|
|
|
|
|
|
def __init__(self, connection_alias, **connection_options):
|
|
|
"""
|
|
|
初始化后端实例
|
|
|
:param connection_alias: settings 中 HAYSTACK_CONNECTIONS 的 key
|
|
|
:param connection_options: 该连接的配置字典
|
|
|
"""
|
|
|
super(WhooshSearchBackend, self).__init__(connection_alias, **connection_options)
|
|
|
self.setup_complete = False # 延迟 setup 标记
|
|
|
self.use_file_storage = True # 默认使用文件存储
|
|
|
# 提交缓冲区大小，默认 128 MB
|
|
|
self.post_limit = getattr(connection_options, 'POST_LIMIT', 128 * 1024 * 1024)
|
|
|
self.path = connection_options.get('PATH') # 索引存放路径
|
|
|
|
|
|
# 如果显式指定 STORAGE != 'file'，则使用内存存储
|
|
|
if connection_options.get('STORAGE', 'file') != 'file':
|
|
|
self.use_file_storage = False
|
|
|
|
|
|
# 文件存储模式下 PATH 不能为空
|
|
|
if self.use_file_storage and not self.path:
|
|
|
raise ImproperlyConfigured(
|
|
|
"You must specify a 'PATH' in your settings for connection '%s'." % connection_alias)
|
|
|
|
|
|
self.log = logging.getLogger('haystack')
|
|
|
|
|
|
def setup(self):
|
|
|
"""
|
|
|
延迟初始化：真正用到时才创建/打开索引，避免进程启动时即锁定索引。
|
|
|
"""
|
|
|
from haystack import connections
|
|
|
new_index = False
|
|
|
|
|
|
# 文件存储：目录不存在则创建
|
|
|
if self.use_file_storage and not os.path.exists(self.path):
|
|
|
os.makedirs(self.path)
|
|
|
new_index = True
|
|
|
|
|
|
# 确保目录可写
|
|
|
if self.use_file_storage and not os.access(self.path, os.W_OK):
|
|
|
raise IOError(
|
|
|
"The path to your Whoosh index '%s' is not writable for the current user/group." % self.path)
|
|
|
|
|
|
# 根据配置选择存储后端
|
|
|
if self.use_file_storage:
|
|
|
self.storage = FileStorage(self.path)
|
|
|
else:
|
|
|
global LOCALS
|
|
|
# 内存模式：线程本地共享 RamStorage
|
|
|
if getattr(LOCALS, 'RAM_STORE', None) is None:
|
|
|
LOCALS.RAM_STORE = RamStorage()
|
|
|
self.storage = LOCALS.RAM_STORE
|
|
|
|
|
|
# 构建 whoosh Schema
|
|
|
self.content_field_name, self.schema = self.build_schema(
|
|
|
connections[self.connection_alias].get_unified_index().all_searchfields())
|
|
|
self.parser = QueryParser(self.content_field_name, schema=self.schema)
|
|
|
|
|
|
# 创建或打开索引
|
|
|
if new_index is True:
|
|
|
self.index = self.storage.create_index(self.schema)
|
|
|
else:
|
|
|
try:
|
|
|
self.index = self.storage.open_index(schema=self.schema)
|
|
|
except index.EmptyIndexError:
|
|
|
self.index = self.storage.create_index(self.schema)
|
|
|
|
|
|
self.setup_complete = True
|
|
|
|
|
|
def build_schema(self, fields):
|
|
|
"""
|
|
|
把 Haystack 的 SearchField 列表转换成 whoosh 的 Schema
|
|
|
:param fields: dict {field_name: field_instance}
|
|
|
:return: (content_field_name, Schema)
|
|
|
"""
|
|
|
schema_fields = {
|
|
|
ID: WHOOSH_ID(stored=True, unique=True), # 主键
|
|
|
DJANGO_CT: WHOOSH_ID(stored=True), # 模型类名
|
|
|
DJANGO_ID: WHOOSH_ID(stored=True), # 模型 pk
|
|
|
}
|
|
|
initial_key_count = len(schema_fields)
|
|
|
content_field_name = ''
|
|
|
|
|
|
for field_name, field_class in fields.items():
|
|
|
if field_class.is_multivalued:
|
|
|
# 多值字段：非索引用 IDLIST，索引用 KEYWORD
|
|
|
if field_class.indexed is False:
|
|
|
schema_fields[field_class.index_fieldname] = IDLIST(
|
|
|
stored=True, field_boost=field_class.boost)
|
|
|
else:
|
|
|
schema_fields[field_class.index_fieldname] = KEYWORD(
|
|
|
stored=True, commas=True, scorable=True, field_boost=field_class.boost)
|
|
|
elif field_class.field_type in ['date', 'datetime']:
|
|
|
schema_fields[field_class.index_fieldname] = DATETIME(
|
|
|
stored=field_class.stored, sortable=True)
|
|
|
elif field_class.field_type == 'integer':
|
|
|
schema_fields[field_class.index_fieldname] = NUMERIC(
|
|
|
stored=field_class.stored, numtype=int, field_boost=field_class.boost)
|
|
|
elif field_class.field_type == 'float':
|
|
|
schema_fields[field_class.index_fieldname] = NUMERIC(
|
|
|
stored=field_class.stored, numtype=float, field_boost=field_class.boost)
|
|
|
elif field_class.field_type == 'boolean':
|
|
|
# BOOLEAN 不支持 field_boost
|
|
|
schema_fields[field_class.index_fieldname] = BOOLEAN(
|
|
|
stored=field_class.stored)
|
|
|
elif field_class.field_type == 'ngram':
|
|
|
schema_fields[field_class.index_fieldname] = NGRAM(
|
|
|
minsize=3, maxsize=15, stored=field_class.stored, field_boost=field_class.boost)
|
|
|
elif field_class.field_type == 'edge_ngram':
|
|
|
schema_fields[field_class.index_fieldname] = NGRAMWORDS(
|
|
|
minsize=2, maxsize=15, at='start',
|
|
|
stored=field_class.stored, field_boost=field_class.boost)
|
|
|
else:
|
|
|
# 默认 TEXT，使用 jieba 中文分词
|
|
|
schema_fields[field_class.index_fieldname] = TEXT(
|
|
|
stored=True, analyzer=ChineseAnalyzer(),
|
|
|
field_boost=field_class.boost, sortable=True)
|
|
|
|
|
|
# 标记文档主字段（用于高亮/拼写检查）
|
|
|
if field_class.document is True:
|
|
|
content_field_name = field_class.index_fieldname
|
|
|
schema_fields[field_class.index_fieldname].spelling = True
|
|
|
|
|
|
# 没有任何业务字段则抛错
|
|
|
if len(schema_fields) <= initial_key_count:
|
|
|
raise SearchBackendError(
|
|
|
"No fields were found in any search_indexes. Please correct this before attempting to search.")
|
|
|
|
|
|
return (content_field_name, Schema(**schema_fields))
|
|
|
|
|
|
def update(self, index, iterable, commit=True):
|
|
|
"""
|
|
|
批量更新/新增文档
|
|
|
:param index: SearchIndex 实例
|
|
|
:param iterable: 要索引的模型实例可迭代对象
|
|
|
:param commit: 是否立即提交（whoosh 2.5+ 建议写完即 commit）
|
|
|
"""
|
|
|
if not self.setup_complete:
|
|
|
self.setup()
|
|
|
|
|
|
self.index = self.index.refresh()
|
|
|
writer = AsyncWriter(self.index) # 异步写，避免长期锁
|
|
|
|
|
|
for obj in iterable:
|
|
|
try:
|
|
|
doc = index.full_prepare(obj) # 提取字段值
|
|
|
except SkipDocument:
|
|
|
self.log.debug(u"Indexing for object `%s` skipped", obj)
|
|
|
continue
|
|
|
|
|
|
# 全部转 unicode
|
|
|
for key in doc:
|
|
|
doc[key] = self._from_python(doc[key])
|
|
|
|
|
|
# whoosh 2.5+ 已移除文档级 boost
|
|
|
if 'boost' in doc:
|
|
|
del doc['boost']
|
|
|
|
|
|
try:
|
|
|
writer.update_document(**doc) # 存在即更新
|
|
|
except Exception as e:
|
|
|
if not self.silently_fail:
|
|
|
raise
|
|
|
self.log.error(
|
|
|
u"%s while preparing object for update" % e.__class__.__name__,
|
|
|
exc_info=True,
|
|
|
extra={"data": {"index": index, "object": get_identifier(obj)}})
|
|
|
|
|
|
if len(iterable) > 0:
|
|
|
writer.commit() # 真正落地
|
|
|
|
|
|
def remove(self, obj_or_string, commit=True):
|
|
|
"""
|
|
|
根据唯一标识删除单条文档
|
|
|
"""
|
|
|
if not self.setup_complete:
|
|
|
self.setup()
|
|
|
|
|
|
self.index = self.index.refresh()
|
|
|
whoosh_id = get_identifier(obj_or_string)
|
|
|
|
|
|
try:
|
|
|
self.index.delete_by_query(
|
|
|
q=self.parser.parse(u'%s:"%s"' % (ID, whoosh_id)))
|
|
|
except Exception as e:
|
|
|
if not self.silently_fail:
|
|
|
raise
|
|
|
self.log.error(
|
|
|
"Failed to remove document '%s' from Whoosh: %s",
|
|
|
whoosh_id, e, exc_info=True)
|
|
|
|
|
|
def clear(self, models=None, commit=True):
|
|
|
"""
|
|
|
清空整个索引或指定模型的索引
|
|
|
:param models: None 表示全部；list/tuple 表示指定模型
|
|
|
"""
|
|
|
if not self.setup_complete:
|
|
|
self.setup()
|
|
|
|
|
|
self.index = self.index.refresh()
|
|
|
|
|
|
if models is not None:
|
|
|
assert isinstance(models, (list, tuple))
|
|
|
|
|
|
try:
|
|
|
if models is None:
|
|
|
self.delete_index() # 删目录/清内存
|
|
|
else:
|
|
|
models_to_delete = []
|
|
|
for model in models:
|
|
|
models_to_delete.append(
|
|
|
u"%s:%s" % (DJANGO_CT, get_model_ct(model)))
|
|
|
# 构造 OR 查询一次性删除
|
|
|
self.index.delete_by_query(
|
|
|
q=self.parser.parse(u" OR ".join(models_to_delete)))
|
|
|
except Exception as e:
|
|
|
if not self.silently_fail:
|
|
|
raise
|
|
|
if models is not None:
|
|
|
self.log.error(
|
|
|
"Failed to clear Whoosh index of models '%s': %s",
|
|
|
','.join(models_to_delete), e, exc_info=True)
|
|
|
else:
|
|
|
self.log.error("Failed to clear Whoosh index: %s", e, exc_info=True)
|
|
|
|
|
|
def delete_index(self):
|
|
|
"""
|
|
|
物理删除索引文件/内存，并重新 setup
|
|
|
"""
|
|
|
if self.use_file_storage and os.path.exists(self.path):
|
|
|
shutil.rmtree(self.path)
|
|
|
elif not self.use_file_storage:
|
|
|
self.storage.clean()
|
|
|
self.setup()
|
|
|
|
|
|
def optimize(self):
|
|
|
"""
|
|
|
手动合并索引段，提升查询速度（耗时操作）
|
|
|
"""
|
|
|
if not self.setup_complete:
|
|
|
self.setup()
|
|
|
self.index = self.index.refresh()
|
|
|
self.index.optimize()
|
|
|
|
|
|
def calculate_page(self, start_offset=0, end_offset=None):
|
|
|
"""
|
|
|
把 Django 风格的分页起止偏移换算成 whoosh 的页码+每页条数
|
|
|
"""
|
|
|
if end_offset is not None and end_offset <= 0:
|
|
|
end_offset = 1
|
|
|
page_num = 0
|
|
|
if end_offset is None:
|
|
|
end_offset = 1000000
|
|
|
if start_offset is None:
|
|
|
start_offset = 0
|
|
|
page_length = end_offset - start_offset
|
|
|
if page_length > 0:
|
|
|
page_num = int(start_offset / page_length)
|
|
|
# whoosh 页码从 1 开始
|
|
|
page_num += 1
|
|
|
return page_num, page_length
|
|
|
|
|
|
@log_query
|
|
|
def search(self, query_string, sort_by=None, start_offset=0, end_offset=None,
|
|
|
fields='', highlight=False, facets=None, date_facets=None,
|
|
|
query_facets=None, narrow_queries=None, spelling_query=None,
|
|
|
within=None, dwithin=None, distance_point=None, models=None,
|
|
|
limit_to_registered_models=None, result_class=None, **kwargs):
|
|
|
"""
|
|
|
核心查询方法，返回 {'results': [...], 'hits': n, ...}
|
|
|
各参数含义见 Haystack 文档，此处不赘述
|
|
|
"""
|
|
|
if not self.setup_complete:
|
|
|
self.setup()
|
|
|
|
|
|
# 空查询直接返回 0 条
|
|
|
if len(query_string) == 0:
|
|
|
return {'results': [], 'hits': 0}
|
|
|
|
|
|
query_string = force_str(query_string)
|
|
|
|
|
|
# 单个非通配字符会被 whoosh 当 stopword 过滤掉，直接返回 0 条
|
|
|
if len(query_string) <= 1 and query_string != u'*':
|
|
|
return {'results': [], 'hits': 0}
|
|
|
|
|
|
reverse = False
|
|
|
# 处理排序方向（whoosh 要求所有字段同向）
|
|
|
if sort_by is not None:
|
|
|
sort_by_list = []
|
|
|
reverse_counter = 0
|
|
|
for order_by in sort_by:
|
|
|
if order_by.startswith('-'):
|
|
|
reverse_counter += 1
|
|
|
if reverse_counter and reverse_counter != len(sort_by):
|
|
|
raise SearchBackendError("Whoosh requires all order_by fields"
|
|
|
" to use the same sort direction")
|
|
|
for order_by in sort_by:
|
|
|
if order_by.startswith('-'):
|
|
|
sort_by_list.append(order_by[1:])
|
|
|
if len(sort_by_list) == 1:
|
|
|
reverse = True
|
|
|
else:
|
|
|
sort_by_list.append(order_by)
|
|
|
if len(sort_by_list) == 1:
|
|
|
reverse = False
|
|
|
sort_by = sort_by_list[0]
|
|
|
|
|
|
# whoosh 不支持 faceting，仅警告
|
|
|
if facets is not None:
|
|
|
warnings.warn("Whoosh does not handle faceting.", Warning, stacklevel=2)
|
|
|
if date_facets is not None:
|
|
|
warnings.warn("Whoosh does not handle date faceting.", Warning, stacklevel=2)
|
|
|
if query_facets is not None:
|
|
|
warnings.warn("Whoosh does not handle query faceting.", Warning, stacklevel=2)
|
|
|
|
|
|
narrowed_results = None
|
|
|
self.index = self.index.refresh()
|
|
|
|
|
|
# 模型过滤
|
|
|
if limit_to_registered_models is None:
|
|
|
limit_to_registered_models = getattr(
|
|
|
settings, 'HAYSTACK_LIMIT_TO_REGISTERED_MODELS', True)
|
|
|
|
|
|
if models and len(models):
|
|
|
model_choices = sorted(get_model_ct(model) for model in models)
|
|
|
elif limit_to_registered_models:
|
|
|
model_choices = self.build_models_list()
|
|
|
else:
|
|
|
model_choices = []
|
|
|
|
|
|
if len(model_choices) > 0:
|
|
|
if narrow_queries is None:
|
|
|
narrow_queries = set()
|
|
|
narrow_queries.add(' OR '.join(
|
|
|
['%s:%s' % (DJANGO_CT, rm) for rm in model_choices]))
|
|
|
|
|
|
narrow_searcher = None
|
|
|
if narrow_queries is not None:
|
|
|
# 用 searcher 先过滤缩小结果集
|
|
|
narrow_searcher = self.index.searcher()
|
|
|
for nq in narrow_queries:
|
|
|
recent_narrowed_results = narrow_searcher.search(
|
|
|
self.parser.parse(force_str(nq)), limit=None)
|
|
|
if len(recent_narrowed_results) <= 0:
|
|
|
return {'results': [], 'hits': 0}
|
|
|
if narrowed_results:
|
|
|
narrowed_results.filter(recent_narrowed_results)
|
|
|
else:
|
|
|
narrowed_results = recent_narrowed_results
|
|
|
|
|
|
self.index = self.index.refresh()
|
|
|
|
|
|
if self.index.doc_count():
|
|
|
searcher = self.index.searcher()
|
|
|
parsed_query = self.parser.parse(query_string)
|
|
|
if parsed_query is None: # 非法/全停词查询
|
|
|
return {'results': [], 'hits': 0}
|
|
|
|
|
|
page_num, page_length = self.calculate_page(start_offset, end_offset)
|
|
|
search_kwargs = {'pagelen': page_length,
|
|
|
'sortedby': sort_by, 'reverse': reverse}
|
|
|
if narrowed_results is not None:
|
|
|
search_kwargs['filter'] = narrowed_results
|
|
|
|
|
|
try:
|
|
|
raw_page = searcher.search_page(parsed_query, page_num, **search_kwargs)
|
|
|
except ValueError:
|
|
|
if not self.silently_fail:
|
|
|
raise
|
|
|
return {'results': [], 'hits': 0, 'spelling_suggestion': None}
|
|
|
|
|
|
# whoosh 2.5.1 在请求页码过大时会返回错误页，需检测
|
|
|
if raw_page.pagenum < page_num:
|
|
|
return {'results': [], 'hits': 0, 'spelling_suggestion': None}
|
|
|
|
|
|
results = self._process_results(
|
|
|
raw_page, highlight=highlight, query_string=query_string,
|
|
|
spelling_query=spelling_query, result_class=result_class)
|
|
|
searcher.close()
|
|
|
if hasattr(narrow_searcher, 'close'):
|
|
|
narrow_searcher.close()
|
|
|
return results
|
|
|
else:
|
|
|
# 索引为空时返回拼写建议
|
|
|
if self.include_spelling:
|
|
|
if spelling_query:
|
|
|
spelling_suggestion = self.create_spelling_suggestion(spelling_query)
|
|
|
else:
|
|
|
spelling_suggestion = self.create_spelling_suggestion(query_string)
|
|
|
else:
|
|
|
spelling_suggestion = None
|
|
|
return {'results': [], 'hits': 0, 'spelling_suggestion': spelling_suggestion}
|
|
|
|
|
|
def more_like_this(self, model_instance, additional_query_string=None,
|
|
|
start_offset=0, end_offset=None, models=None,
|
|
|
limit_to_registered_models=None, result_class=None, **kwargs):
|
|
|
"""
|
|
|
根据给定实例找“相似文档”
|
|
|
"""
|
|
|
if not self.setup_complete:
|
|
|
self.setup()
|
|
|
|
|
|
# 处理延迟加载模型
|
|
|
model_klass = model_instance._meta.concrete_model
|
|
|
field_name = self.content_field_name
|
|
|
narrow_queries = set()
|
|
|
narrowed_results = None
|
|
|
self.index = self.index.refresh()
|
|
|
|
|
|
if limit_to_registered_models is None:
|
|
|
limit_to_registered_models = getattr(
|
|
|
settings, 'HAYSTACK_LIMIT_TO_REGISTERED_MODELS', True)
|
|
|
|
|
|
if models and len(models):
|
|
|
model_choices = sorted(get_model_ct(model) for model in models)
|
|
|
elif limit_to_registered_models:
|
|
|
model_choices = self.build_models_list()
|
|
|
else:
|
|
|
model_choices = []
|
|
|
|
|
|
if len(model_choices) > 0:
|
|
|
if narrow_queries is None:
|
|
|
narrow_queries = set()
|
|
|
narrow_queries.add(' OR '.join(
|
|
|
['%s:%s' % (DJANGO_CT, rm) for rm in model_choices]))
|
|
|
|
|
|
if additional_query_string and additional_query_string != '*':
|
|
|
narrow_queries.add(additional_query_string)
|
|
|
|
|
|
narrow_searcher = None
|
|
|
if narrow_queries is not None:
|
|
|
narrow_searcher = self.index.searcher()
|
|
|
for nq in narrow_queries:
|
|
|
recent_narrowed_results = narrow_searcher.search(
|
|
|
self.parser.parse(force_str(nq)), limit=None)
|
|
|
if len(recent_narrowed_results) <= 0:
|
|
|
return {'results': [], 'hits': 0}
|
|
|
if narrowed_results:
|
|
|
narrowed_results.filter(recent_narrowed_results)
|
|
|
else:
|
|
|
narrowed_results = recent_narrowed_results
|
|
|
|
|
|
page_num, page_length = self.calculate_page(start_offset, end_offset)
|
|
|
self.index = self.index.refresh()
|
|
|
raw_results = EmptyResults()
|
|
|
|
|
|
if self.index.doc_count():
|
|
|
query = "%s:%s" % (ID, get_identifier(model_instance))
|
|
|
searcher = self.index.searcher()
|
|
|
parsed_query = self.parser.parse(query)
|
|
|
results = searcher.search(parsed_query)
|
|
|
if len(results):
|
|
|
# 取第一条结果调 more_like_this
|
|
|
raw_results = results[0].more_like_this(
|
|
|
field_name, top=end_offset)
|
|
|
if narrowed_results is not None and hasattr(raw_results, 'filter'):
|
|
|
raw_results.filter(narrowed_results)
|
|
|
|
|
|
try:
|
|
|
raw_page = ResultsPage(raw_results, page_num, page_length)
|
|
|
except ValueError:
|
|
|
if not self.silently_fail:
|
|
|
raise
|
|
|
return {'results': [], 'hits': 0, 'spelling_suggestion': None}
|
|
|
|
|
|
if raw_page.pagenum < page_num:
|
|
|
return {'results': [], 'hits': 0, 'spelling_suggestion': None}
|
|
|
|
|
|
results = self._process_results(raw_page, result_class=result_class)
|
|
|
searcher.close()
|
|
|
if hasattr(narrow_searcher, 'close'):
|
|
|
narrow_searcher.close()
|
|
|
return results
|
|
|
|
|
|
def _process_results(self, raw_page, highlight=False, query_string='',
|
|
|
spelling_query=None, result_class=None):
|
|
|
"""
|
|
|
把 whoosh 的 ResultsPage 转成 haystack SearchResult 列表
|
|
|
"""
|
|
|
from haystack import connections
|
|
|
results = []
|
|
|
hits = len(raw_page) # 必须在切片前取总数
|
|
|
|
|
|
if result_class is None:
|
|
|
result_class = SearchResult
|
|
|
|
|
|
facets = {}
|
|
|
spelling_suggestion = None
|
|
|
unified_index = connections[self.connection_alias].get_unified_index()
|
|
|
indexed_models = unified_index.get_indexed_models()
|
|
|
|
|
|
for doc_offset, raw_result in enumerate(raw_page):
|
|
|
score = raw_page.score(doc_offset) or 0
|
|
|
app_label, model_name = raw_result[DJANGO_CT].split('.')
|
|
|
additional_fields = {}
|
|
|
model = haystack_get_model(app_label, model_name)
|
|
|
|
|
|
if model and model in indexed_models:
|
|
|
for key, value in raw_result.items():
|
|
|
index = unified_index.get_index(model)
|
|
|
string_key = str(key)
|
|
|
if string_key in index.fields and hasattr(index.fields[string_key], 'convert'):
|
|
|
# 多值 KEYWORD 字段用逗号拆分
|
|
|
if index.fields[string_key].is_multivalued:
|
|
|
if value is None or len(value) == 0:
|
|
|
additional_fields[string_key] = []
|
|
|
else:
|
|
|
additional_fields[string_key] = value.split(',')
|
|
|
else:
|
|
|
additional_fields[string_key] = index.fields[string_key].convert(value)
|
|
|
else:
|
|
|
additional_fields[string_key] = self._to_python(value)
|
|
|
|
|
|
# 删除内部字段
|
|
|
del additional_fields[DJANGO_CT]
|
|
|
del additional_fields[DJANGO_ID]
|
|
|
|
|
|
# 高亮处理
|
|
|
if highlight:
|
|
|
sa = StemmingAnalyzer()
|
|
|
formatter = WhooshHtmlFormatter('em')
|
|
|
terms = [token.text for token in sa(query_string)]
|
|
|
whoosh_result = whoosh_highlight(
|
|
|
additional_fields.get(self.content_field_name),
|
|
|
terms, sa, ContextFragmenter(), formatter)
|
|
|
additional_fields['highlighted'] = {
|
|
|
self.content_field_name: [whoosh_result],
|
|
|
}
|
|
|
|
|
|
result = result_class(app_label, model_name,
|
|
|
raw_result[DJANGO_ID], score,
|
|
|
**additional_fields)
|
|
|
results.append(result)
|
|
|
else:
|
|
|
hits -= 1 # 模型未注册，命中不计入总数
|
|
|
|
|
|
# 拼写建议
|
|
|
if self.include_spelling:
|
|
|
if spelling_query:
|
|
|
spelling_suggestion = self.create_spelling_suggestion(spelling_query)
|
|
|
else:
|
|
|
spelling_suggestion = self.create_spelling_suggestion(query_string)
|
|
|
|
|
|
return {'results': results, 'hits': hits,
|
|
|
'facets': facets, 'spelling_suggestion': spelling_suggestion}
|
|
|
|
|
|
def create_spelling_suggestion(self, query_string):
|
|
|
"""
|
|
|
基于 whoosh corrector 给出拼写纠正建议
|
|
|
"""
|
|
|
spelling_suggestion = None
|
|
|
reader = self.index.reader()
|
|
|
corrector = reader.corrector(self.content_field_name)
|
|
|
cleaned_query = force_str(query_string)
|
|
|
if not query_string:
|
|
|
return spelling_suggestion
|
|
|
# 去掉保留字/符
|
|
|
for rev_word in self.RESERVED_WORDS:
|
|
|
cleaned_query = cleaned_query.replace(rev_word, '')
|
|
|
for rev_char in self.RESERVED_CHARACTERS:
|
|
|
cleaned_query = cleaned_query.replace(rev_char, '')
|
|
|
# 按单词纠正
|
|
|
query_words = cleaned_query.split()
|
|
|
suggested_words = []
|
|
|
for word in query_words:
|
|
|
suggestions = corrector.suggest(word, limit=1)
|
|
|
if len(suggestions) > 0:
|
|
|
suggested_words.append(suggestions[0])
|
|
|
spelling_suggestion = ' '.join(suggested_words)
|
|
|
return spelling_suggestion
|
|
|
|
|
|
def _from_python(self, value):
|
|
|
"""
|
|
|
把 Python 值转成 whoosh 可索引的 unicode / string
|
|
|
"""
|
|
|
if hasattr(value, 'strftime'):
|
|
|
if not hasattr(value, 'hour'):
|
|
|
value = datetime(value.year, value.month, value.day, 0, 0, 0)
|
|
|
elif isinstance(value, bool):
|
|
|
value = 'true' if value else 'false'
|
|
|
elif isinstance(value, (list, tuple)):
|
|
|
value = u','.join([force_str(v) for v in value])
|
|
|
elif isinstance(value, (six.integer_types, float)):
|
|
|
pass # 数字保持原样
|
|
|
else:
|
|
|
value = force_str(value)
|
|
|
return value
|
|
|
|
|
|
def _to_python(self, value):
|
|
|
"""
|
|
|
把 whoosh 存储的字符串转回 Python 对象
|
|
|
"""
|
|
|
if value == 'true':
|
|
|
return True
|
|
|
elif value == 'false':
|
|
|
return False
|
|
|
if value and isinstance(value, six.string_types):
|
|
|
possible_datetime = DATETIME_REGEX.search(value)
|
|
|
if possible_datetime:
|
|
|
date_values = possible_datetime.groupdict()
|
|
|
for dk, dv in date_values.items():
|
|
|
date_values[dk] = int(dv)
|
|
|
return datetime(date_values['year'], date_values['month'],
|
|
|
date_values['day'], date_values['hour'],
|
|
|
date_values['minute'], date_values['second'])
|
|
|
try:
|
|
|
converted_value = json.loads(value)
|
|
|
if isinstance(converted_value, (list, tuple, set, dict,
|
|
|
six.integer_types, float, complex)):
|
|
|
return converted_value
|
|
|
except BaseException:
|
|
|
pass
|
|
|
return value
|
|
|
|
|
|
|
|
|
class WhooshSearchQuery(BaseSearchQuery):
|
|
|
"""
|
|
|
负责把 Haystack 的 filter/exclude 条件转换成 whoosh 查询语法
|
|
|
"""
|
|
|
|
|
|
def _convert_datetime(self, date):
|
|
|
"""
|
|
|
统一把 datetime 转成 whoosh 需要的 YYYYMMDDHHMMSS 字符串
|
|
|
"""
|
|
|
if hasattr(date, 'hour'):
|
|
|
return force_str(date.strftime('%Y%m%d%H%M%S'))
|
|
|
else:
|
|
|
return force_str(date.strftime('%Y%m%d000000'))
|
|
|
|
|
|
def clean(self, query_fragment):
|
|
|
"""
|
|
|
转义保留字/符；whoosh 1.x 不再支持反斜杠转义，需用引号包裹
|
|
|
"""
|
|
|
words = query_fragment.split()
|
|
|
cleaned_words = []
|
|
|
for word in words:
|
|
|
if word in self.backend.RESERVED_WORDS:
|
|
|
word = word.lower() # 保留字小写化
|
|
|
for char in self.backend.RESERVED_CHARACTERS:
|
|
|
if char in word:
|
|
|
word = "'%s'" % word
|
|
|
break
|
|
|
cleaned_words.append(word)
|
|
|
return ' '.join(cleaned_words)
|
|
|
|
|
|
def build_query_fragment(self, field, filter_type, value):
|
|
|
"""
|
|
|
把单个 filter 条件转成 whoosh 查询子串
|
|
|
例如：name='exact' -> name:xxx
|
|
|
"""
|
|
|
from haystack import connections
|
|
|
query_frag = ''
|
|
|
is_datetime = False
|
|
|
|
|
|
if not hasattr(value, 'input_type_name'):
|
|
|
# 处理 ValuesListQuerySet 等
|
|
|
if hasattr(value, 'values_list'):
|
|
|
value = list(value)
|
|
|
if hasattr(value, 'strftime'):
|
|
|
is_datetime = True
|
|
|
if isinstance(value, six.string_types) and value != ' ':
|
|
|
value = Clean(value)
|
|
|
else:
|
|
|
value = PythonData(value)
|
|
|
|
|
|
prepared_value = value.prepare(self)
|
|
|
|
|
|
if not isinstance(prepared_value, (set, list, tuple)):
|
|
|
prepared_value = self.backend._from_python(prepared_value)
|
|
|
|
|
|
# 'content' 是保留字段，表示全文检索
|
|
|
if field == 'content':
|
|
|
index_fieldname = ''
|
|
|
else:
|
|
|
index_fieldname = u'%s:' % connections[self._using].get_unified_index(
|
|
|
).get_index_fieldname(field)
|
|
|
|
|
|
# 查询模板映射
|
|
|
filter_types = {
|
|
|
'content': '%s',
|
|
|
'contains': '*%s*',
|
|
|
'endswith': "*%s",
|
|
|
'startswith': "%s*",
|
|
|
'exact': '%s',
|
|
|
'gt': "{%s to}",
|
|
|
'gte': "[%s to]",
|
|
|
'lt': "{to %s}",
|
|
|
'lte': "[to %s]",
|
|
|
'fuzzy': u'%s~',
|
|
|
}
|
|
|
|
|
|
if value.post_process is False:
|
|
|
query_frag = prepared_value
|
|
|
else:
|
|
|
if filter_type in ['content', 'contains', 'startswith',
|
|
|
'endswith', 'fuzzy']:
|
|
|
if value.input_type_name == 'exact':
|
|
|
query_frag = prepared_value
|
|
|
else:
|
|
|
# 分词后每个词都加通配符/模板
|
|
|
terms = []
|
|
|
if isinstance(prepared_value, six.string_types):
|
|
|
possible_values = prepared_value.split(' ')
|
|
|
else:
|
|
|
if is_datetime is True:
|
|
|
prepared_value = self._convert_datetime(prepared_value)
|
|
|
possible_values = [prepared_value]
|
|
|
for possible_value in possible_values:
|
|
|
terms.append(
|
|
|
filter_types[filter_type] % self.backend._from_python(possible_value))
|
|
|
if len(terms) == 1:
|
|
|
query_frag = terms[0]
|
|
|
else:
|
|
|
query_frag = u"(%s)" % " AND ".join(terms)
|
|
|
elif filter_type == 'in':
|
|
|
in_options = []
|
|
|
for possible_value in prepared_value:
|
|
|
is_datetime = False
|
|
|
if hasattr(possible_value, 'strftime'):
|
|
|
is_datetime = True
|
|
|
pv = self.backend._from_python(possible_value)
|
|
|
if is_datetime is True:
|
|
|
pv = self._convert_datetime(pv)
|
|
|
if isinstance(pv, six.string_types) and not is_datetime:
|
|
|
in_options.append('"%s"' % pv)
|
|
|
else:
|
|
|
in_options.append('%s' % pv)
|
|
|
query_frag = "(%s)" % " OR ".join(in_options)
|
|
|
elif filter_type == 'range':
|
|
|
start = self.backend._from_python(prepared_value[0])
|
|
|
end = self.backend._from_python(prepared_value[1])
|
|
|
if hasattr(prepared_value[0], 'strftime'):
|
|
|
start = self._convert_datetime(start)
|
|
|
if hasattr(prepared_value[1], 'strftime'):
|
|
|
end = self._convert_datetime(end)
|
|
|
query_frag = u"[%s to %s]" % (start, end)
|
|
|
elif filter_type == 'exact':
|
|
|
if value.input_type_name == 'exact':
|
|
|
query_frag = prepared_value
|
|
|
else:
|
|
|
prepared_value = Exact(prepared_value).prepare(self)
|
|
|
query_frag = filter_types[filter_type] % prepared_value
|
|
|
else:
|
|
|
if is_datetime is True:
|
|
|
prepared_value = self._convert_datetime(prepared_value)
|
|
|
query_frag = filter_types[filter_type] % prepared_value
|
|
|
|
|
|
# 非 Raw 输入且未带括号，则整体括号包裹
|
|
|
if len(query_frag) and not isinstance(value, Raw):
|
|
|
if not query_frag.startswith('(') and not query_frag.endswith(')'):
|
|
|
query_frag = "(%s)" % query_frag
|
|
|
|
|
|
return u"%s%s" % (index_fieldname, query_frag)
|
|
|
|
|
|
|
|
|
class WhooshEngine(BaseEngine):
|
|
|
"""
|
|
|
入口 Engine，供 Haystack 加载
|
|
|
"""
|
|
|
backend = WhooshSearchBackend
|
|
|
query = WhooshSearchQuery |
