为什么原子操作被认为是线程安全的？(Why are atomic operations considered thread-safe?)

 这里有一些巧妙的技巧。 考虑这个增强模型： 
 
class Person(db.Model):
  first_name = db.StringProperty()
  last_name = db.StringProperty()
  middle_name = db.StringProperty()
  names_lower = db.StringListProperty()
 
 您需要让names_lower与真实字段保持同步，例如： 
 
p.names_lower = [p.first_name.lower(), p.last_name.lower(),
                 p.middle_name.lower()]
 
 您可以使用DerivedProperty更优雅地完成此操作 。 
 
 现在，您的查询： 
 
term = self.request.get('term').lower()
query = Person.all()
query.filter('names_lower >=', term)
query.filter('names_lower <=', unicode(term) + u"\ufffd")
 
 这给你： 
 
 
 
 使用一个索引匹配所有3个属性 
 
 
 不区分大小写的匹配项 
 
 
 通配符后缀匹配 
 
 
 因此，对“smi”的查询将返回任何名称以“smi”开头的人。 
 
 将较小的名称复制到ListProperty可启用不区分大小写的匹配，并允许我们使用一个查询搜索所有3个字段。 “\ ufffd”是最高可能的unicode字符，因此它是我们子字符串匹配的上限。 如果由于某种原因您想要完全匹配，请过滤'names_lower =', term而不是'names_lower =', term 。 
 
 编辑 ： 
 
 
 
 我应该如何在我的数据存储区中搜索相同的内容（因为我需要查看每个字段，可能还有3个字段的组合值）？ 如何优化这样的查询？ 
 
 
 这已经在原始解决方案中得到了解决。 通过获取3个字段并将它们复制到单个ListProperty，我们实际上创建了一个单个索引，每个人有多个条目。 如果我们有一个名叫Bob J Smith的人，我的索引中会有3个点击： 
 
 
 
 names_lower = bob 
 
 
 names_lower = j 
 
 
 names_lower =史密斯 
 
 
 这消除了在每个字段上运行不同查询的需要。 
 
 
 
 我也不清楚答复格式应该是什么。 
 
 
 仔细阅读文档 。 格式化jQuery的输出应该非常简单。 您的数据源将是指定URL的字符串，您需要将响应格式化为JSON。 

There are a few neat tricks here. Consider this augmented model:
 
class Person(db.Model):
  first_name = db.StringProperty()
  last_name = db.StringProperty()
  middle_name = db.StringProperty()
  names_lower = db.StringListProperty()
 
You'll need to keep names_lower in sync with the real fields, e.g.:
 
p.names_lower = [p.first_name.lower(), p.last_name.lower(),
                 p.middle_name.lower()]
 
You can do this more elegantly with a DerivedProperty.
 
And now, your query:
 
term = self.request.get('term').lower()
query = Person.all()
query.filter('names_lower >=', term)
query.filter('names_lower <=', unicode(term) + u"\ufffd")
 
This gives you:
 
 
 
Matching on all 3 properties with one index
 
 
Case insensitive matches
 
 
Wildcard suffix matches
 
 
So a query for "smi" will return any person with any name starting with "smi" in any case.
 
Copying lower-cased names to a ListProperty enables case-insensitive matching, and allows us to search all 3 fields with one query. "\ufffd" is the highest possible unicode character, so it's the upper limit for our substring match. If for some reason you want an exact match, filter for 'names_lower =', term instead.
 
Edit:
 
 
 
How should I search for the same in my datastore (since I need to look at each field and probably for combined value of 3 fields)? How to optimize such query?
 
 
This is already accounted for in the original solution. By taking the 3 fields and copying them to a single ListProperty, we're essentially creating a single index with multiple entries per person. If we have a person named Bob J Smith, he'll have 3 hits in our index:
 
 
 
names_lower = bob
 
 
names_lower = j
 
 
names_lower = smith
 
 
This eliminates the need to run distinct queries on each field.
 
 
 
I am also not clear what should be the reply format.
 
 
Read the docs carefully. Formatting output for jQuery should be pretty straightforward. Your data source will be a string specifying a URL, and you'll want to format the response as JSON.