Docs Menu
Docs Home
/
MongoDB Manual
/
Aggregation Operations
/
Reference
/
Operators

$dateSubtract (aggregation)

On this page

  • Definition
  • Behavior
  • Examples

Definition

$dateSubtract

New in version 5.0.

Decrements a Date() object by a specified number of time units.

The $dateSubtract expression has the following syntax:

{
   $dateSubtract: {
      startDate: <Expression>,
      unit: <Expression>,
      amount: <Expression>,
      timezone: <tzExpression>
   }
}

Returns a Date(). The startDate can be any expression that resolves to type Date, Timestamp or ObjectId. No matter which data type is used as input, the value returned will be a Date() object.

Field
Required/Optional
Description
startDate
Required
The beginning date, in UTC, for the subtraction operation. The startDate can be any expression that resolves to a Date, a Timestamp, or an ObjectID.
unit
Required

The unit used to measure the amount of time subtracted from the startDate. The unit is an expression that resolves to one of the following strings:

  • year

  • quarter

  • week

  • month

  • day

  • hour

  • minute

  • second

  • millisecond

amount
Required
The number of units subtracted from the startDate. The amount is an expression that resolves to an integer or long. The amount can also resolve to an integral decimal and or a double if that value can be converted to a long without loss of precision.
timezone
Optional

The timezone to carry out the operation. <tzExpression> must be a valid expression that resolves to a string formatted as either an Olson Timezone Identifier or a UTC Offset. If no timezone is provided, the result is displayed in UTC.

Format
Examples
Olson Timezone Identifier
"America/New_York"
"Europe/London"
"GMT"
UTC Offset
+/-[hh]:[mm], e.g. "+04:45"
+/-[hh][mm], e.g. "-0530"
+/-[hh], e.g. "+03"

For more information on expressions and types see Expression Operators and BSON Types.

Behavior

Time Measurement

MongoDB follows prevaling database usage and works with time in UTC. The dateSubtract expression always takes a startDate in UTC and returns a result in UTC. If the timezone is specified, the calculation will be done using the specified timezone. The timezone is especially important when a calculation involves Daylight Savings Time (DST).

If the unit is a month, or larger the operation adjusts to account for the last day of the month. Subtracting one month on the last day of March, for example, demonstrates the "last-day-of-the-month" adjustment.

{
   $dateSubtract:
      {
         startDate: ISODate("2021-03-31T12:10:05Z"),
         unit: "month",
         amount: 1
      }
}

Notice that the date returned, ISODate("2021-02-28T12:10:05Z"), is the 28th and not the 31st since February has fewer days than March.

Time Zone

When using an Olson Timezone Identifier in the <timezone> field, MongoDB applies the DST offset if applicable for the specified timezone.

For example, consider a sales collection with the following document:

{
   "_id" : 1,
   "item" : "abc",
   "price" : 20,
   "quantity" : 5,
   "date" : ISODate("2017-05-20T10:24:51.303Z")
}

The following aggregation illustrates how MongoDB handles the DST offset for the Olson Timezone Identifier. The example uses the $hour and $minute operators to return the corresponding portions of the date field:

db.sales.aggregate([
{
   $project: {
      "nycHour": {
         $hour: { date: "$date", timezone: "-05:00" }
       },
       "nycMinute": {
          $minute: { date: "$date", timezone: "-05:00" }
       },
       "gmtHour": {
          $hour: { date: "$date", timezone: "GMT" }
       },
       "gmtMinute": {
          $minute: { date: "$date", timezone: "GMT" } },
       "nycOlsonHour": {
          $hour: { date: "$date", timezone: "America/New_York" }
       },
       "nycOlsonMinute": {
          $minute: { date: "$date", timezone: "America/New_York" }
       }
   }
}])

The operation returns the following result:

{
   "_id": 1,
   "nycHour" : 5,
   "nycMinute" : 24,
   "gmtHour" : 10,
   "gmtMinute" : 24,
   "nycOlsonHour" : 6,
   "nycOlsonMinute" : 24
}

Examples

Subtract A Fixed Amount

Consider a collection of system connection times like these:

db.connectionTime.insertMany(
  [
     {
        custId: 457,
        login: ISODate("2020-12-25T19:04:00"),
        logout: ISODate("2020-12-28T09:04:00")
     },
     {
        custId: 457,
        login: ISODate("2021-01-27T05:12:00"),
        logout: ISODate("2021-01-28T13:05:00")
     },
     {
        custId: 458,
        login: ISODate("2021-01-22T06:27:00"),
        logout: ISODate("2021-01-31T11:00:00")
     },
     {
        custId: 459,
        login: ISODate("2021-02-14T20:14:00"),
        logout: ISODate("2021-02-17T16:05:00")
     },
     {
        custId: 460,
        login: ISODate("2021-02-26T02:44:00"),
        logout: ISODate("2021-02-18T14:13:00")
        }
  ]
)

Due to a service issue you need to subtract 3 hours from each of the January 2021 logout times. You can use $dateSubtract in an aggregation pipeline to decrement the logoutTime.

db.connectionTime.aggregate(
   [
      {
         $match:
            {
               $expr:
                  {
                     $eq:
                        [
                            { $year: "$logout" },
                              2021
                        ]
                  },
               $expr:
                  {
                     $eq:
                        [
                            { $month: "$logout" },
                              1
                        ]
                   }
             }
       },
       {
          $project:
             {
                logoutTime:
                   {
                      $dateSubtract:
                         {
                            startDate: "$logout",
                            unit: "hour",
                            amount: 3
                         }
                   }
              }
        },
        {
           $merge: "connectionTime"
        }
   ]
)

Two similar comparisons are made in the $match stage. First the $year and $month operators extract the year and month, respectively, from the logoutTime Date object. Then the month and year are checked to see if they match the selection targets. Since "January" is encoded as "1", $expr is true when the year and month are equal ($eq) to "2021" and "1".

The $project stage uses $dateSubtract to subtract 3 hours from the logoutTime of each selected dcoument.

Finaly, the $merge stage updates the collection, writing the new logoutTime for the modified documents.

Note

Unlike $out, the $merge stage only updates the matched documents and preserves the rest of the collection. For more details see: $out compared with $merge.

The resulting documents look like this:

{
 "_id" : ObjectId("603dd94b044b995ad331c0b5"),
 "custId" : 457,
 "login" : ISODate("2020-12-25T19:04:00Z"),
 "logout" : ISODate("2020-12-28T09:04:00Z")
}
{
 "_id" : ObjectId("603dd94b044b995ad331c0b6"),
 "custId" : 457,
 "login" : ISODate("2021-01-27T05:12:00Z"),
 "logout" : ISODate("2021-01-28T13:05:00Z"),
 "logoutTime" : ISODate("2021-01-28T10:05:00Z")
}
{
 "_id" : ObjectId("603dd94b044b995ad331c0b7"),
 "custId" : 458,
 "login" : ISODate("2021-01-22T06:27:00Z"),
 "logout" : ISODate("2021-01-31T11:00:00Z"),
 "logoutTime" : ISODate("2021-01-31T08:00:00Z")
}
{
 "_id" : ObjectId("603dd94b044b995ad331c0b8"),
 "custId" : 459,
 "login" : ISODate("2021-02-14T20:14:00Z"),
 "logout" : ISODate("2021-02-17T16:05:00Z")
}
{
 "_id" : ObjectId("603dd94b044b995ad331c0b9"),
 "custId" : 460,
 "login" : ISODate("2021-02-26T02:44:00Z"),
 "logout" : ISODate("2021-02-18T14:13:00Z")
}

Filter by Relative Dates

You want to send a survey to clients who have used your service in the past week. The $dateSubtract expression can create a range filter relative to the time the query is executed.

db.connectionTime.aggregate(
   [
      {
         $match:
            {
               $expr:
                  {
                     $gt:
                        [
                           "$logoutTime",
                            {
                               $dateSubtract:
                                  {
                                     startDate: "$$NOW",
                                     unit: "week",
                                     amount: 1
                                  }
                            }
                        ]
                  }
             }
      },
      {
         $project:
            {
               _id: 0,
               custId: 1,
               loggedOut:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d",
                           date: "$logoutTime"
                        }
                  }
            }
      }
   ]
)

The built in aggregation variable $$NOW returns the current datetime in ISODate format. The $match stage uses the value in $$NOW to get today's date. Then the comparison expression ($expr) filters the collection using greater than ($gt) and $dateSubtract to match documents that have a logoutTime in the past week.

The $project stage uses the $dateToString expression to convert the dates to a more readable format. Without the conversion, MongoDB returns the date in ISODate format and assumes a UTC timezone.

The output shows two customers have logged out in the last week.

{ "custId" : 459, "loggedOut" : "2021-02-17" }
{ "custId" : 460, "loggedOut" : "2021-02-18" }

Adjust for Daylight Savings Time

All dates are stored internally in UTC time. When a timezone is specified, $dateSubtract uses local time to carry out the calculations. The results are displayed in UTC.

You have customers in several timezones and you want to see what effect daylight savings time might have on your billing periods if you bill by day or by hour.

Create this collection of connection times:

db.billing.insertMany(
   [
      {
         location: "America/New_York",
         login: ISODate("2021-03-14T10:00:00-0500"),
         logout: ISODate("2021-03-14T18:00:00-0500")
      },
      {
         location: "America/Mexico_City",
         login: ISODate("2021-03-14T10:00:00-00:00"),
         logout: ISODate("2021-03-15T08:00:00-0500")
      }
   ]
)

First subtract 1 day, then subtract 24 hours from the login dates in each document.

db.billing.aggregate(
   [
      {
         $project:
            {
               _id: 0,
               location: 1,
               start:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d %H:%M",
                           date: "$login"
                        }
                  },
               days:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d %H:%M",
                           date:
                              {
                                 $dateSubtract:
                                    {
                                       startDate: "$login",
                                       unit: "day",
                                       amount: 1,
                                       timezone: "$location"
                                    }
                              }
                        }
                  },
               hours:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d %H:%M",
                           date:
                              {
                                 $dateSubtract:
                                 {
                                    startDate: "$login",
                                    unit: "hour",
                                    amount: 24,
                                    timezone: "$location"
                                 }
                              }
                        }
                  },
               startTZInfo:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d %H:%M",
                           date: "$login",
                           timezone: "$location"
                        }
                  },
               daysTZInfo:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d %H:%M",
                           date:
                              {
                                 $dateSubtract:
                                    {
                                       startDate: "$login",
                                       unit: "day",
                                       amount: 1,
                                       timezone: "$location"
                                    }
                              },
                           timezone: "$location"
                        }
                  },
               hoursTZInfo:
                  {
                     $dateToString:
                        {
                           format: "%Y-%m-%d %H:%M",
                           date:
                              {
                                 $dateSubtract:
                                    {
                                       startDate: "$login",
                                       unit: "hour",
                                       amount: 24,
                                       timezone: "$location"
                                    }
                              },
                           timezone: "$location"
                        }
                  },
            }
      }
   ]
).pretty()

The $dateToString expression reformats the output for readability. Results are summarized here:

Field
New York
Mexico City
Start
2021-03-14 15:00
2021-03-14 15:00
Start, TZ Info
2021-03-14 11:00
2021-03-14 04:00
1 Day
2021-03-13 16:00
2021-03-13 15:00
1 Day, TZInfo
2021-03-13 11:00
2021-03-13 09:00
24 Hours
2021-03-13 15:00
2021-03-13 15:00
24 Hours, TZInfo
2021-03-13 10:00
2021-03-13 09:00

The chart highlights several points:

  • Unformatted dates are returned in UTC. The $login for New York is UTC -5, however the start, days, and hours rows display the time in UTC.

  • March 14th is the start of DST in New York, but not in Mexico. The calculated time is adjusted when a location switches to DST and crosses from one day to the next.

  • DST modifies the length of the day, not the hour. There is no DST change for hours. There is an only an adjustment for DST when the measurement unit is day or larger and the computation crosses a clock change in the specified timezone.

Tip

See also:

$dateAdd, $dateDiff

Back

$dateFromString