How to query Azure CLI command output using a JMESPath query
The Azure CLI uses the --query parameter to execute a JMESPath query on the results of commands. JMESPath is a query language for JSON, giving you the ability to select and modify data from CLI output.
All commands in Azure CLI support the --query parameter. This article covers how to use the features of JMESPath and gives examples of queries. Learn about JMESPath concepts that are useful for querying under the concepts tab. See examples of JMESPath queries under the examples tab.
Azure CLI uses queries to select and modify the output of Azure CLI commands. Queries are executed client-side on the Azure CLI command's returned JSON object before any display formatting.
The escape characters needed in queries differ for different environments. It's recommended to run queries in Azure Cloud Shell or cmd because these shells require fewer escape characters. To ensure the query examples are syntactically correct, select the tab for the shell you're using.
Dictionary and list CLI results
CLI command results are first treated as JSON for queries, even when the output format is something other than JSON. CLI results are either a JSON array or dictionary. Arrays are sequences of objects that can be indexed, and dictionaries are unordered objects accessed with keys.
This is an example of an array:
[
1,
2,
3
]
This is an example of a dictionary:
{
"isRunning": false,
"time": "12:00",
"number": 1
}
Commands that could return more than one object return an array, and commands that always return only a single object return a dictionary.
Get properties in a dictionary
Working with dictionary results, you can access properties from the top level with just the key. The . (subexpression) character is used to access properties of nested dictionaries. Before introducing queries, take a look at the unmodified output of the az vm show command:
az vm show --resource-group QueryDemo --name TestVM
The command outputs a dictionary. Some content has been omitted.
{
"additionalCapabilities": null,
"availabilitySet": null,
"diagnosticsProfile": {
"bootDiagnostics": {
"enabled": true,
"storageUri": "https://xxxxxx.blob.core.windows.net/"
}
},
...
"osProfile": {
"adminPassword": null,
"adminUsername": "azureuser",
"allowExtensionOperations": true,
"computerName": "TestVM",
"customData": null,
"linuxConfiguration": {
"disablePasswordAuthentication": true,
"provisionVmAgent": true,
"ssh": {
"publicKeys": [
{
"keyData": "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDMobZNJTqgjWn/IB5xlilvE4Y+BMYpqkDnGRUcA0g9BYPgrGSQquCES37v2e3JmpfDPHFsaR+CPKlVr2GoVJMMHeRcMJhj50ZWq0hAnkJBhlZVWy8S7dwdGAqPyPmWM2iJDCVMVrLITAJCno47O4Ees7RCH6ku7kU86b1NOanvrNwqTHr14wtnLhgZ0gQ5GV1oLWvMEVg1YFMIgPRkTsSQKWCG5lLqQ45aU/4NMJoUxGyJTL9i8YxMavaB1Z2npfTQDQo9+womZ7SXzHaIWC858gWNl9e5UFyHDnTEDc14hKkf1CqnGJVcCJkmSfmrrHk/CkmF0ZT3whTHO1DhJTtV stramer@contoso",
"path": "/home/azureuser/.ssh/authorized_keys"
}
]
}
},
"secrets": [],
"windowsConfiguration": null
},
....
}
The following command gets the SSH public keys authorized to connect to the VM by adding a query:
az vm show --resource-group QueryDemo --name TestVM --query "osProfile.linuxConfiguration.ssh.publicKeys"
[
{
"keyData": "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDMobZNJTqgjWn/IB5xlilvE4Y+BMYpqkDnGRUcA0g9BYPgrGSQquCES37v2e3JmpfDPHFsaR+CPKlVr2GoVJMMHeRcMJhj50ZWq0hAnkJBhlZVWy8S7dwdGAqPyPmWM2iJDCVMVrLITAJCno47O4Ees7RCH6ku7kU86b1NOanvrNwqTHr14wtnLhgZ0gQ5GV1oLWvMEVg1YFMIgPRkTsSQKWCG5lLqQ45aU/4NMJoUxGyJTL9i8YxMavaB1Z2npfTQDQo9+womZ7SXzHaIWC858gWNl9e5UFyHDnTEDc14hKkf1CqnGJVcCJkmSfmrrHk/CkmF0ZT3whTHO1DhJTtV stramer@contoso",
"path": "/home/azureuser/.ssh/authorized_keys"
}
]
Query strings are case sensitive. For example, changing 'osProfile' to 'OsProfile' in the previous query doesn't return the correct results.
Get multiple values
To get more than one property, put expressions separated by commas in square brackets [ ] (a multiselect list). The following command gets the VM name, admin user, and SSH key all at once:
az vm show --resource-group QueryDemo --name TestVM --query "[name, osProfile.adminUsername, osProfile.linuxConfiguration.ssh.publicKeys[0].keyData]"
[
"TestVM",
"azureuser",
"ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDMobZNJTqgjWn/IB5xlilvE4Y+BMYpqkDnGRUcA0g9BYPgrGSQquCES37v2e3JmpfDPHFsaR+CPKlVr2GoVJMMHeRcMJhj50ZWq0hAnkJBhlZVWy8S7dwdGAqPyPmWM2iJDCVMVrLITAJCno47O4Ees7RCH6ku7kU86b1NOanvrNwqTHr14wtnLhgZ0gQ5GV1oLWvMEVg1YFMIgPRkTsSQKWCG5lLqQ45aU/4NMJoUxGyJTL9i8YxMavaB1Z2npfTQDQo9+womZ7SXzHaIWC858gWNl9e5UFyHDnTEDc14hKkf1CqnGJVcCJkmSfmrrHk/CkmF0ZT3whTHO1DhJTtV stramer@contoso"
]
These values are listed in the result array in the order they were given in the query. Since the result is an array, there are no keys associated with the results. To get a dictionary instead of an array, see the next section.
Rename properties in a query
To get a dictionary instead of an array when querying for multiple values, use the { } (multiselect hash) operator.
The format for a multiselect hash is {displayName:JMESPathExpression, ...}.
displayName is the string shown in output, and JMESPathExpression is the JMESPath expression to evaluate. Modify the example from the
last section by changing the multiselect list to a hash:
Note
If you choose to use a space in a new column name, like VM name instead of VMName, the quoting
rules change in both Bash and PowerShell. See Pass spaces in Azure CLI parameters
for examples.
az vm show --resource-group QueryDemo --name TestVM --query "{VMName:name, admin:osProfile.adminUsername, sshKey:osProfile.linuxConfiguration.ssh.publicKeys[0].keyData}"
{
"VMName": "TestVM",
"admin": "azureuser",
"ssh-key": "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDMobZNJTqgjWn/IB5xlilvE4Y+BMYpqkDnGRUcA0g9BYPgrGSQquCES37v2e3JmpfDPHFsaR+CPKlVr2GoVJMMHeRcMJhj50ZWq0hAnkJBhlZVWy8S7dwdGAqPyPmWM2iJDCVMVrLITAJCno47O4Ees7RCH6ku7kU86b1NOanvrNwqTHr14wtnLhgZ0gQ5GV1oLWvMEVg1YFMIgPRkTsSQKWCG5lLqQ45aU/4NMJoUxGyJTL9i8YxMavaB1Z2npfTQDQo9+womZ7SXzHaIWC858gWNl9e5UFyHDnTEDc14hKkf1CqnGJVcCJkmSfmrrHk/CkmF0ZT3whTHO1DhJTtV stramer@contoso"
}
Get properties in an array
An array has no properties of its own, but it can be indexed. This feature is shown in the last example with the expression publicKeys[0], which gets the first element of the publicKeys array. There's no guarantee CLI output is ordered, so avoid using indexing unless you're sure of the order or don't care which element you get. To access the properties of elements in an array, you do one of two operations: flattening or filtering. This section covers how to flatten an array.
Flattening an array is done with the [] JMESPath operator. All expressions after the [] operator are applied to each element in the current array.
If [] appears at the start of the query, it flattens the CLI command result. The results of az vm list can be inspected with this feature.
The following query gets the name, OS, and administrator name for each VM in a resource group:
az vm list --resource-group QueryDemo --query "[].{Name:name, OS:storageProfile.osDisk.osType, admin:osProfile.adminUsername}"
[
{
"Name": "Test-2",
"OS": "Linux",
"admin": "sttramer"
},
{
"Name": "TestVM",
"OS": "Linux",
"admin": "azureuser"
},
{
"Name": "WinTest",
"OS": "Windows",
"admin": "winadmin"
}
]
Any array can be flattened, not just the top-level result returned by the command. In the last section, the expression osProfile.linuxConfiguration.ssh.publicKeys[0].keyData was used to get the SSH public key for sign-in. To get every SSH public key, the expression could instead be written as osProfile.linuxConfiguration.ssh.publicKeys[].keyData.
This query expression flattens the osProfile.linuxConfiguration.ssh.publicKeys array, and then runs the keyData expression on each
element:
az vm show --resource-group QueryDemo --name TestVM --query "{VMName:name, admin:osProfile.adminUsername, sshKeys:osProfile.linuxConfiguration.ssh.publicKeys[].keyData }"
{
"VMName": "TestVM",
"admin": "azureuser",
"sshKeys": [
"ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDMobZNJTqgjWn/IB5xlilvE4Y+BMYpqkDnGRUcA0g9BYPgrGSQquCES37v2e3JmpfDPHFsaR+CPKlVr2GoVJMMHeRcMJhj50ZWq0hAnkJBhlZVWy8S7dwdGAqPyPmWM2iJDCVMVrLITAJCno47O4Ees7RCH6ku7kU86b1NOanvrNwqTHr14wtnLhgZ0gQ5GV1oLWvMEVg1YFMIgPRkTsSQKWCG5lLqQ45aU/4NMJoUxGyJTL9i8YxMavaB1Z2npfTQDQo9+womZ7SXzHaIWC858gWNl9e5UFyHDnTEDc14hKkf1CqnGJVcCJkmSfmrrHk/CkmF0ZT3whTHO1DhJTtV stramer@contoso\n"
]
}
Filter arrays with boolean expressions
The other operation used to get data from an array is filtering. Filtering is done with the [?...] JMESPath operator.
This operator takes a predicate as its contents. A predicate is any statement (including Boolean properties) that can be evaluated to either true or false. Expressions where the predicate evaluates to true are included in the output.
The first query demonstrates how to list the names of all Azure subscriptions connected to your account whose isDefault property is true. The second and third queries show two different ways to list all subscriptions whose isDefault property is false.
# Boolean values are assumed to be true, so you can directly evaluate the isDefault property to return the default subscription.
az account list --query "[?isDefault].name"
# To check if a Boolean property is false, you can use the comparison operator == or the logical operator !.
az account list --query '[?!isDefault].name'
az account list --query "[?isDefault == \`false\`].name"
JMESPath offers the standard comparison and logical operators. These include <, <=, >, >=, ==, and !=. JMESPath also supports logical and (&&), or (||), and not (!). Expressions can be grouped within parenthesis, allowing for more complex predicate expressions. For the full details on predicates and logical operations, see the JMESPath specification.
In the last section, you flattened an array to get the complete list of all VMs in a resource group. With the use of filters, this output can be restricted to only Linux VMs:
az vm list --resource-group QueryDemo --query "[?storageProfile.osDisk.osType=='Linux'].{Name:name, admin:osProfile.adminUsername}" --output table
Name Admin
------ ---------
Test-2 sttramer
TestVM azureuser
You can also filter numerical values such as the OS disk size. The following example demonstrates how to filter the list of VMs to display ones with a disk size larger than or equal to 50 GB.
az vm list --resource-group QueryDemo --query "[?storageProfile.osDisk.diskSizeGb >=\`50\`].{Name:name, admin:osProfile.adminUsername, DiskSize:storageProfile.osDisk.diskSizeGb }" --output table
Name Admin DiskSize
------- -------- --------
WinTest winadmin 127
For large arrays, it may be faster to apply the filter before selecting data.
Important
In JMESPath, strings are always surrounded by single quotes (') or escape characters (`). If you use double quotes as part of a string in a filter predicate, you'll get empty output.
JMESPath functions
JMESPath also has built-in functions that allow for more complex queries and for modifying query output. This section focuses on using JMESPath functions to create queries while the Manipulating output with functions section demonstrates how to use functions to modify the output.
Expressions are evaluated before calling the function, so arguments themselves can be JMESPath expressions. The following examples demonstrate this concept by using contains(string, substring), which checks to see if a string contains a substring. This command finds all VMs using SSD storage for their OS disk:
az vm list --resource-group QueryDemo --query "[?contains(storageProfile.osDisk.managedDisk.storageAccountType,'SSD')].{Name:name, Storage:storageProfile.osDisk.managedDisk.storageAccountType}"
[
{
"Name": "TestVM",
"Storage": "StandardSSD_LRS"
},
{
"Name": "WinTest",
"Storage": "StandardSSD_LRS"
}
]
Pipe expressions
Similar to how | is used in the command line, | can be used in JMESPath queries to apply expressions to intermediate query results. We can also use | to break down complex queries into simpler subexpressions. To shorten the query from the previous section, use | to apply the filter after flattening and selecting data.
az vm list --resource-group QueryDemo --query "[].{Name:name, Storage:storageProfile.osDisk.managedDisk.storageAccountType} | [? contains(Storage,'SSD')]"
[
{
"Name": "TestVM",
"Storage": "StandardSSD_LRS"
},
{
"Name": "WinTest",
"Storage": "StandardSSD_LRS"
}
]
See the JMESPath specification - Built-in Functions for the full list of functions.
Manipulating output with functions
JMESPath functions also have another purpose, which is to operate on the results of a query. Any function that returns a nonboolean value changes the result of an expression. For example, you can sort data by a property value with sort_by(array, &sort_expression). JMESPath uses a special operator, &, for expressions that should be evaluated later as part of a function. The next example shows how to sort a VM list by OS disk size:
az vm list --resource-group QueryDemo --query "sort_by([].{Name:name, Size:storageProfile.osDisk.diskSizeGb}, &Size)" --output table
Name Size
------- ------
Test-2 30
TestVM 32
WinTest 127
See the JMESPath specification - Built-in Functions for the full list of functions.
Formatting query results
The Azure CLI uses JSON as its default output format, however different output formats may better suit a query depending on its purpose and results. Queries are always run on the JSON output first and then formatted.
This section will go over tsv and table formatting and some use cases for each format. For more information about output formats, see Output formats for Azure CLI commands.
TSV output format
The tsv output format returns tab- and newline-separated values without extra formatting, keys, or other symbols. This format is useful when the output is stored in a parameter and used in another command.
One use case for tsv formatting is queries that retrieve a value out of a CLI command, such as an Azure resource ID or resource name, and store the value in a local environment variable. By default the results are returned in JSON format, which may be an issue when dealing with JSON strings that are enclosed in " characters. The quotes may not be interpreted by the shell if the command output is directly assigned to the environment variable. This issue be seen in the following example that assigns a query result to an environment variable:
USER=$(az vm show --resource-group QueryDemo --name TestVM --query "osProfile.adminUsername")
echo $USER
"azureuser"
Use tsv formatting, as demonstrated in the following query, to prevent enclosing return values with type information:
USER=$(az vm show --resource-group QueryDemo --name TestVM --query "osProfile.adminUsername" --output tsv)
echo $USER
azureuser
Table output format
The table format prints output as an ASCII table, making it easy to read and scan. Not all fields are included in the table so this format is best used as a human-searchable overview of data. Fields that aren't included in the table can still be filtered for as part of a query.
Note
Certain keys are filtered out and not printed in the table view. These keys are id, type, and etag. To see these values, you can change the key name in a multiselect hash.
az vm show --resource-group QueryDemo --name TestVM --query "{objectID:id}" --output table
We can use a previous query to demonstrate this concept. The original query returned a JSON object containing the name, OS, and administrator name for each VM in the resource group:
az vm list --resource-group QueryDemo --query "[].{Name:name, OS:storageProfile.osDisk.osType, admin:osProfile.adminUsername}"
[
{
"Name": "Test-2",
"OS": "Linux",
"admin": "sttramer"
},
{
"Name": "TestVM",
"OS": "Linux",
"admin": "azureuser"
},
{
"Name": "WinTest",
"OS": "Windows",
"admin": "winadmin"
}
]
When combined with the --output table output format, the column names match up with the displayKey value of the multiselect hash making it easier to skim the information:
az vm list --resource-group QueryDemo --query "[].{Name:name, OS:storageProfile.osDisk.osType, Admin:osProfile.adminUsername}" --output table
Name OS Admin
------- ------- ---------
Test-2 Linux sttramer
TestVM Linux azureuser
WinTest Windows winadmin
Next steps
To learn more about JMESPath queries, see JMESPath Tutorial.
To learn more about other Azure CLI concepts mentioned in this article see:
Feedback
Coming soon: Throughout 2024 we will be phasing out GitHub Issues as the feedback mechanism for content and replacing it with a new feedback system. For more information see: https://aka.ms/ContentUserFeedback.
Submit and view feedback for