I'm thinking that UTC time strings like 2011-01-26 21:41:09 +0000 might be okay since they sort correctly if they are used in a view key, but storing the time zone (e.g. 2011-01-26 16:41:09 -0500) would make the document more readable. Converting the date into an epoch integer seem the least appealing from a readability standpoint, but maybe best for performance (or does it make a difference?). What's the recommended practice here?
我认为UTC时间字符串(2011-01-26 21:41:09 +0000)可能是可以的，因为如果在视图键中使用的话，它们可以进行正确的排序，但是存储时区(例如2011-01-26 16:41:09 -0500)将使文档更具可读性。从可读性的角度来看，将日期转换为一个历元整数似乎是最不吸引人的，但也许这对性能来说是最好的(或者这对性能有影响吗?)这里推荐的练习是什么?
5 个解决方案
#1
29
Time is a one-dimensional thing. A timestamp plus timezone is two-dimensional, describing a point in time, and a location. Couch views are one-dimensional (but not the GeoCouch plugin), so storing in a common zone (UTC) is wise.
时间是一维的。时间戳加上时区是二维的，描述时间点和位置。沙发视图是一维的(但不是GeoCouch插件)，所以在公共区域(UTC)中存储是明智的。
Probably the most future-proof format is a string that naturally sorts in chronological order. Probably the most convenient such format is what JSON2 outputs:
可能最能证明未来的格式是按时间顺序自然排序的字符串。可能最方便的这种格式是JSON2输出的:
> a = new Date();
Thu Jan 27 2011 18:40:52 GMT+0700 (ICT)
> JSON.stringify(a)
"2011-01-27T11:40:52.280Z"
#2
6
If you're just using the Map side of Map reduce than these suggestions are probably fine. If, however, you want to do a reduce on the results (_count, _stats, _sum), then I'd recommend emitting your dates as arrays so you can use group_level.
如果你只是使用地图的地图一面减少，这些建议可能是好的。但是，如果您希望对结果(_count、_stats、_sum)进行减少，那么我建议将日期作为数组发送，以便您可以使用group_level。
For instance, if you emit(doc.date.split('-')) on a date strings formatted like "2011-02-14", then you could return _count's (for instance) per day, month, and year by using group_level=3, 2, and 1 respectively.
例如，如果您在日期字符串(如“2011-02-14”)中发出(doc.date.split('-'))，那么您可以使用group_level=3、2和1分别返回_count(例如)每天、月和年。
You can further filter the data by adding non-date data to the beginning of the key. If you were outputting Twitter names, for instance, your key might look like ["bigbluehat", "2011", "02", "14"] and your reduce could return the total count of all tweets for the user "bigbluehat" as well as stats for that user across day, month, and year.
您可以通过在键的开头添加非日期数据来进一步过滤数据。例如，如果您输出Twitter名称，您的键可能看起来像["bigbluehat"、"2011"、"02"、"14"]，您的reduce可以返回用户"bigbluehat"的所有tweet总数，以及该用户的每日、每月和年度统计数据。
If you're not using the reduce side of things, then string-based keys are likely fine.
如果您不使用reduce部分，那么基于字符串的键很可能没问题。
#3
5
No matter what kind of data storage I use, I typically want a unix timestamp in there as a field, in which I'll include one for the created date, and then an updated field that I can change when the document changes.
无论我使用哪种类型的数据存储，我通常都希望在其中有一个unix时间戳作为一个字段，其中包含一个用于创建日期的字段，然后是一个更新的字段，当文档发生更改时，可以更改该字段。
I prefer the regular "seconds since epoch" approach rather than "milliseconds since epoch" simply for brevety.
对于brevety，我更喜欢常规的“秒since epoch”方法，而不是“毫秒since epoch”。
Math.round(new Date().getTime()/1000) does the trick for me.
数学。round(new Date(). gettime()/1000)对我来说很有用。
In terms of readibility, i want to store it as an integer for easy comparisons, and use the front end to display it nicely.
至于易读性，我想将它存储为整数，以便于比较，并使用前端来很好地显示它。
#4
4
You can store your dates how ever you want*, it is how you output them into your views that is important.
你可以将你的约会日期存储到你想要的位置，重要的是你如何将它们输出到你的视图中。
*As long as Date.parse() can read it.
只要Date.parse()可以读取它。
There is a good solution here: Sorting Dates in CouchDB Views
这里有一个很好的解决方案:在CouchDB视图中排序日期
#5
3
I like to use milliseconds since last epoch. You can figure this out with:
我喜欢从上一个纪元开始使用毫秒。你可以用:
new Date().valueOf()
You can create a new date from milliseconds with:
你可以用毫秒来创建一个新的日期:
var milliseconds = new Date().valueOf();
var date = new Date(milliseconds);
I like to create a view where the timestamp (in milliseconds) is the key b/c sorting is super-easy that way.
我喜欢创建一个视图，其中时间戳(以毫秒为单位)是关键的b/c排序方式非常简单。
Also, I think using integers is more efficient than strings, at least when it comes to working with data outside of CouchDB.
而且，我认为使用整数比字符串更有效，至少在处理CouchDB之外的数据时是这样。
#1
29
Time is a one-dimensional thing. A timestamp plus timezone is two-dimensional, describing a point in time, and a location. Couch views are one-dimensional (but not the GeoCouch plugin), so storing in a common zone (UTC) is wise.
时间是一维的。时间戳加上时区是二维的，描述时间点和位置。沙发视图是一维的(但不是GeoCouch插件)，所以在公共区域(UTC)中存储是明智的。
Probably the most future-proof format is a string that naturally sorts in chronological order. Probably the most convenient such format is what JSON2 outputs:
可能最能证明未来的格式是按时间顺序自然排序的字符串。可能最方便的这种格式是JSON2输出的:
> a = new Date();
Thu Jan 27 2011 18:40:52 GMT+0700 (ICT)
> JSON.stringify(a)
"2011-01-27T11:40:52.280Z"
#2
6
If you're just using the Map side of Map reduce than these suggestions are probably fine. If, however, you want to do a reduce on the results (_count, _stats, _sum), then I'd recommend emitting your dates as arrays so you can use group_level.
如果你只是使用地图的地图一面减少，这些建议可能是好的。但是，如果您希望对结果(_count、_stats、_sum)进行减少，那么我建议将日期作为数组发送，以便您可以使用group_level。
For instance, if you emit(doc.date.split('-')) on a date strings formatted like "2011-02-14", then you could return _count's (for instance) per day, month, and year by using group_level=3, 2, and 1 respectively.
例如，如果您在日期字符串(如“2011-02-14”)中发出(doc.date.split('-'))，那么您可以使用group_level=3、2和1分别返回_count(例如)每天、月和年。
You can further filter the data by adding non-date data to the beginning of the key. If you were outputting Twitter names, for instance, your key might look like ["bigbluehat", "2011", "02", "14"] and your reduce could return the total count of all tweets for the user "bigbluehat" as well as stats for that user across day, month, and year.
您可以通过在键的开头添加非日期数据来进一步过滤数据。例如，如果您输出Twitter名称，您的键可能看起来像["bigbluehat"、"2011"、"02"、"14"]，您的reduce可以返回用户"bigbluehat"的所有tweet总数，以及该用户的每日、每月和年度统计数据。
If you're not using the reduce side of things, then string-based keys are likely fine.
如果您不使用reduce部分，那么基于字符串的键很可能没问题。
#3
5
No matter what kind of data storage I use, I typically want a unix timestamp in there as a field, in which I'll include one for the created date, and then an updated field that I can change when the document changes.
无论我使用哪种类型的数据存储，我通常都希望在其中有一个unix时间戳作为一个字段，其中包含一个用于创建日期的字段，然后是一个更新的字段，当文档发生更改时，可以更改该字段。
I prefer the regular "seconds since epoch" approach rather than "milliseconds since epoch" simply for brevety.
对于brevety，我更喜欢常规的“秒since epoch”方法，而不是“毫秒since epoch”。
Math.round(new Date().getTime()/1000) does the trick for me.
数学。round(new Date(). gettime()/1000)对我来说很有用。
In terms of readibility, i want to store it as an integer for easy comparisons, and use the front end to display it nicely.
至于易读性，我想将它存储为整数，以便于比较，并使用前端来很好地显示它。
#4
4
You can store your dates how ever you want*, it is how you output them into your views that is important.
你可以将你的约会日期存储到你想要的位置，重要的是你如何将它们输出到你的视图中。
*As long as Date.parse() can read it.
只要Date.parse()可以读取它。
There is a good solution here: Sorting Dates in CouchDB Views
这里有一个很好的解决方案:在CouchDB视图中排序日期
#5
3
I like to use milliseconds since last epoch. You can figure this out with:
我喜欢从上一个纪元开始使用毫秒。你可以用:
new Date().valueOf()
You can create a new date from milliseconds with:
你可以用毫秒来创建一个新的日期:
var milliseconds = new Date().valueOf();
var date = new Date(milliseconds);
I like to create a view where the timestamp (in milliseconds) is the key b/c sorting is super-easy that way.
我喜欢创建一个视图，其中时间戳(以毫秒为单位)是关键的b/c排序方式非常简单。
Also, I think using integers is more efficient than strings, at least when it comes to working with data outside of CouchDB.
而且，我认为使用整数比字符串更有效，至少在处理CouchDB之外的数据时是这样。