Difference between revisions of "MongoDB QueryData"

From mi-linux
Jump to navigationJump to search
(Created page with " Show all data so far: db.deptCollection.find() Comes back messy, can make it better: db.deptCollection.find().pretty() Or just find one document: db.deptCollection.fi...")
 
 
(90 intermediate revisions by the same user not shown)
Line 1: Line 1:
 +
[[Main Page]] >> [[MongoDB|MongoDB]] >>[[MongoDB_Workbook|MongoDB Workbook]] >> Querying Collections
  
 +
== Querying a collection ==
  
Show all data so far:
+
The find() function can be used to query the documents.
db.deptCollection.find()
 
  
Comes back messy, can make it better:
+
The format is:
  
  db.deptCollection.find().pretty()
+
<pre style="color: purple">
 +
  db.collectionName.find(optional_query_criteria)
 +
</pre>
  
Or just find one document:
+
Where the query_criteria follows a pattern:
  
  db.deptCollection.find({"deptno":10}).pretty()
+
<pre style="color: purple">
 +
  db.collectionName.find({keyField: "value"})
 +
</pre>
  
Finding an employee means using the array name too:
+
Note:
db.deptCollection.find({"employees.empno":7902}).pretty()
+
* the criteria is enclosed in curly brackets: {}
 +
* the value needs quotes if it is a string or date value
 +
* all names and values are case sensitive
 +
* quotes are optional for the fieldName, so long as they do not contain spaces
  
However, does mean you get back all the employees!
+
 
 +
=== Find all documents ===
 +
 
 +
For example, show all the data so far in the ''dept'' collection:
 +
db.dept.find()
 +
 
 +
The results should look like:
 +
 
 +
<pre style="color: blue">
 +
{ "_id" : 10, "deptno" : 10, "dname" : "ACCOUNTING", "loc" : "NEW YORK" }
 +
{ "_id" : 20, "deptno" : 20, "dname" : "RESEARCH", "loc" : "DALLAS" }
 +
{ "_id" : 40, "deptno" : 40, "dname" : "OPERATIONS", "loc" : "BOSTON" }
 +
{ "_id" : 50, "deptno" : 40, "dname" : "OPERATIONS V2", "loc" : "BOSTON" }
 +
{ "_id" : 30, "deptno" : 30, "dname" : "SALES", "loc" : "CHICAGO" }
 +
 
 +
</pre>
 +
 
 +
To show the documents in the ''emp'' collection:
 +
 
 +
db.emp.find()
 +
 
 +
 
 +
The data comes back messy. The pretty() function can be used to improve the layout:
 +
 
 +
db.emp.find().pretty()
 +
 
 +
A subset of the ''emp'' collection is shown below:
 +
 
 +
<pre style="color: blue">
 +
> db.emp.find().pretty()
 +
{
 +
        "_id" : ObjectId("5a09e79ac536e890d5a7a666"),
 +
        "empno" : 7782,
 +
        "ename" : "CLARK",
 +
        "job" : "MANAGER",
 +
        "mgr" : 7839,
 +
        "hiredate" : ISODate("1989-06-09T00:00:00Z"),
 +
        "sal" : 2450,
 +
        "deptno" : 10
 +
}
 +
{
 +
        "_id" : ObjectId("5a09e79ac536e890d5a7a667"),
 +
        "empno" : 7839,
 +
        "ename" : "KING",
 +
        "job" : "PRESIDENT",
 +
        "hiredate" : ISODate("1980-11-17T00:00:00Z"),
 +
        "sal" : 5000,
 +
        "deptno" : 10
 +
}
 +
....
 +
 
 +
</pre>
 +
 
 +
Note the object ids are now system generated (and will be different values in your own data).
 +
 
 +
=== Find with query criteria ===
 +
 
 +
 
 +
If working with a large collection, you will not want all the documents returned.
 +
 
 +
Find all the employees are are clerks:
 +
 
 +
  db.emp.find({job:"CLERK"})
 +
 
 +
For numerical data, the greater than (>) and less than (<) operators are represented by ''$gt'' and ''$lt'' respectively. Note, for these operators, the search criteria must be enclosed in {} brackets.
 +
 
 +
Find all employees who earn more than 2400:
 +
db.emp.find({sal: {$gt:2400}})
 +
 
 +
Find all employees whose commission is less than 1000:
 +
db.emp.find({comm: {$lt:1000}})
 +
 
 +
Working with the date field (hiredate) is more complex, since you have to create a new date for the comparison.
 +
 
 +
For example, find all employees who start after the 1st January 2000:
 +
db.emp.find({hiredate: {$lt: new Date("2000-01-01")}})
 +
 
 +
Find employees who started on the 16th October 2015:
 +
db.emp.find({hiredate: new Date("2015-10-16")})
 +
 
 +
=== Find One document ===
 +
 
 +
To find just one document requires the use of the equivalent of a primary key field. This can be a field that the user takes responsibility to keep unique, such as the ''deptno'':
 +
 
 +
db.dept.find({deptno:10})
 +
 
 +
 
 +
Or the object id can be used, which will be unique:
 +
 
 +
db.dept.find({_id:10})
 +
 
 +
 
 +
In the ''emp'' collection, the _ids are system generated and generally along the lines of: '5a0727e99ba81dee9b1cc6a3', so less easy to use!
 +
 
 +
List all the records in emp:
 +
 
 +
db.emp.find().pretty()
 +
 
 +
and pick an _id from the collection and then try and find one record.
 +
 
 +
For example (note, your object id will be different):
 +
 
 +
db.emp.find( {_id : ObjectId("5a0727e99ba81dee9b1cc6a3")}).pretty()
 +
 
 +
 
 +
The function ''ObjectId()'' must be used to convert the value into an object id.
 +
 
 +
== Next Step ==
 +
 
 +
[[MongoDB_Aggregate_Pipeline|MongoDB Aggregation]] Pipeline

Latest revision as of 14:13, 5 March 2018

Main Page >> MongoDB >>MongoDB Workbook >> Querying Collections

Querying a collection

The find() function can be used to query the documents.

The format is:

 db.collectionName.find(optional_query_criteria)

Where the query_criteria follows a pattern:

 db.collectionName.find({keyField: "value"})

Note:

  • the criteria is enclosed in curly brackets: {}
  • the value needs quotes if it is a string or date value
  • all names and values are case sensitive
  • quotes are optional for the fieldName, so long as they do not contain spaces


Find all documents

For example, show all the data so far in the dept collection:

db.dept.find()

The results should look like:

{ "_id" : 10, "deptno" : 10, "dname" : "ACCOUNTING", "loc" : "NEW YORK" }
{ "_id" : 20, "deptno" : 20, "dname" : "RESEARCH", "loc" : "DALLAS" }
{ "_id" : 40, "deptno" : 40, "dname" : "OPERATIONS", "loc" : "BOSTON" }
{ "_id" : 50, "deptno" : 40, "dname" : "OPERATIONS V2", "loc" : "BOSTON" }
{ "_id" : 30, "deptno" : 30, "dname" : "SALES", "loc" : "CHICAGO" }

To show the documents in the emp collection:

db.emp.find()


The data comes back messy. The pretty() function can be used to improve the layout:

db.emp.find().pretty()

A subset of the emp collection is shown below:

> db.emp.find().pretty()
{
        "_id" : ObjectId("5a09e79ac536e890d5a7a666"),
        "empno" : 7782,
        "ename" : "CLARK",
        "job" : "MANAGER",
        "mgr" : 7839,
        "hiredate" : ISODate("1989-06-09T00:00:00Z"),
        "sal" : 2450,
        "deptno" : 10
}
{
        "_id" : ObjectId("5a09e79ac536e890d5a7a667"),
        "empno" : 7839,
        "ename" : "KING",
        "job" : "PRESIDENT",
        "hiredate" : ISODate("1980-11-17T00:00:00Z"),
        "sal" : 5000,
        "deptno" : 10
}
....

Note the object ids are now system generated (and will be different values in your own data).

Find with query criteria

If working with a large collection, you will not want all the documents returned.

Find all the employees are are clerks:

 db.emp.find({job:"CLERK"})

For numerical data, the greater than (>) and less than (<) operators are represented by $gt and $lt respectively. Note, for these operators, the search criteria must be enclosed in {} brackets.

Find all employees who earn more than 2400:

db.emp.find({sal: {$gt:2400}})

Find all employees whose commission is less than 1000:

db.emp.find({comm: {$lt:1000}})

Working with the date field (hiredate) is more complex, since you have to create a new date for the comparison.

For example, find all employees who start after the 1st January 2000:

db.emp.find({hiredate: {$lt: new Date("2000-01-01")}})

Find employees who started on the 16th October 2015:

db.emp.find({hiredate: new Date("2015-10-16")})

Find One document

To find just one document requires the use of the equivalent of a primary key field. This can be a field that the user takes responsibility to keep unique, such as the deptno:

db.dept.find({deptno:10})


Or the object id can be used, which will be unique:

db.dept.find({_id:10})


In the emp collection, the _ids are system generated and generally along the lines of: '5a0727e99ba81dee9b1cc6a3', so less easy to use!

List all the records in emp:

db.emp.find().pretty()

and pick an _id from the collection and then try and find one record.

For example (note, your object id will be different):

db.emp.find( {_id : ObjectId("5a0727e99ba81dee9b1cc6a3")}).pretty()


The function ObjectId() must be used to convert the value into an object id.

Next Step

MongoDB Aggregation Pipeline