How to Sort by Two or More Criteria in Python

The aim of this page📝 is to explain using sorted() with multiple criteria based on the particular example of sorting JSON schema URIs by name and version.

CODE

Let’s jump right into it as the following explanation will make much more sense. Say I have a list of schema endpoints in the following format: <vendor_tag>/<schema_name>/jsonschema/<major>-<minor>-<patch> and I need to sort first by <schema_name> and then by the three last items which together form a version based on semantic versioning. An example of the input is then:

all_schemas = [
    'com.acme/foo/jsonschema/1-0-1',
    'com.acme/example/jsonschema/1-0-0',
    'com.acme/bar/jsonschema/1-0-1',
    'com.acme/foo/jsonschema/1-0-0',
    'com.acme/bar/jsonschema/1-0-0',
]

The function doing the sorting looks as follows:

# The following function returns
def sort(all_schemas: list):
    def get_schema_name(single_schema: str) -> str:
        return single_schema.split('/')[1]

    def get_semver(single_schema: str) -> tuple:
        return tuple(map(int, single_schema.split('/')[-1].split('-')))

    return sorted(all_schemas, key=lambda single_schema: (get_schema_name(single_schema),
                                                          get_semver(single_schema)))

sorted_schemas = sort(all_schemas)

Once called, that function returns exactly what’s required

[
 'com.acme/bar/jsonschema/1-0-0',
 'com.acme/bar/jsonschema/1-0-1',
 'com.acme/example_event/jsonschema/1-0-0'
 'com.acme/foo/jsonschema/1-0-0',
 'com.acme/foo/jsonschema/1-0-1',
]

EXPLAINER

• sorted() function takes a <list> parameter and a key parameter.
• what exactly is sorted() doing is none of our business, sorting is an essential CS subject, I am staying on a primitive “how to use” level without flirting with algorithms here.
• keyis optional and is used for custom sorting
• if you want to sort a list of numbers or words, you don’t really need it.
• key takes only a single parameter that is <callable>- usually a single function
• that single function can however wrap more than one sub-function, which is what we are after if we are to sort by more than 1 criteria
• this function can be a lambda or a regular function
• the lambda can return multiple values that can serve as multiple ordering criteria!
• say you want to first extract by name which is part of a larger structure
• and then, if there are say just 10 namespaces in 1000 rows, you want to sort by something else
• …like a version sequence that is specific within the given namespace
• key function is applied to each element of the input list
• the resulting values are used to compare the elements and determine their order in the sorted list.
• to sort by multiple criteria, the key function can return a tuple containing the values of the criteria that you want to sort by.
• the sorted() function will then sort the list based on the values returned by the key function.

LINKS

• https://docs.python.org/3/howto/sorting.html#sortinghowto