visier-connector 0.9.5

Visier People Data connector

Visier Python Connector

Use the Visier Python Connector to query Visier People data.

The connector enables Python developers to query Visier People data using Visier's SQL-like query language.

Prerequisites

The connector acts as a bridge between your Python application, which is typically Pandas-enabled, and Visier's cloud-hosted service infrastructure. In order to successfully connect to your Visier People data, you need:

• The URL domain name prefix. For example: https://{vanity-name}.api.visier.io.
• An API key issued by Visier.
• A username identifying a user within your organization's Visier tenant who has been granted API access capabilities.
• That user's password

Connector Separation

As of version 0.9.5, the Python connector has separated the API calls from the VisierSession object. As a result of this change, the query execution methods on the VisierSession have been deprecated and will be subject to removal in a future release.

The new way of invoking Visier public APIs through the Visier Python connector requires instantiating the appropriate API client and calling the methods defined on the client object. The following example, invokes the analytic-objects Model API to obtain the metadata for two analytic objects:

    with VisierSession(auth) as session:
        model_client = ModelApiClient(session)

        objs = model_client.get_analytic_objects(["Requisition", "Employee_Exit"])
        print(objs.text)

Error handling

By default, a failed API call will return None and information about the error is available on the client object. Using the example above, the last error in the event objs was None would be model_client.last_error().

It is however possible to force the API client to instead raise a QueryExecutionException. This is accomplished when instantiating the API client with the following parameter value raise_on_error=True. Using the example above, the model_client instantiation would look like this: model_client = ModelApiClient(session, raise_on_error=True).

Examples

Query API

The Query API Client is used to make calls to Visier's Query APIs.

Note that the examples in this repository are not included in the visier-connector package Instead, these examples should be copied into a sample application or the example queries can be run with a test script in this repository as per the snippets below.

The Query API examples use Pandas to illustrate a common data engineering and data science workflow using Visier data.

A small set of example queries have been provided. Generally, Visier Query API queries fall into one of two categories:

1. Detail query - These queries produce tabular results from underlying individual analytic objects. The shape of the result is inherently tabular with each table attribute represented as a column in the result set. Detail queries are often referred to as list or even drill-through queries. This query provides a detailed, non-aggregated view of the underlying analytical objects.
2. Aggregate query - These queries aggregate metric values. They do so along the axes defined for the query and they produce multi-dimensional cell sets by default However, by providing an Accept header whose first value is either application/jsonlines or text/csv, the server will flatten the cell set into a tabular format when building the response.

Visier also offers an experimental alternative to the JSON-based query definitions: SQL-like. This allows you to make queries using a language that comes close to SQL, which is generally more compact and intuitive. SQL-like allows definition of both aggregate and detail queries.

:warning: SQL-like is in alpha stage and not yet suitable for production use.

Example queries are provided through individual files. This is merely for convenience. SQL-like, being simple strings, can easily be provided to the call itself.

In order to reduce duplication, each provided sample below should be preceded by the necessary import statements as well as authentication credential definition:

import os
from visier.connector import Authentication, VisierSession
from visier.api import QueryApiClient
from examples import load_json, load_str
import pandas as pd

auth = Authentication(
    username = os.getenv("SOME_USERNAME"),
    password = os.getenv("SOME_PASSWORD"),
    host = os.getenv("SOME_HOST"),
    api_key = os.getenv("SOME_APIKEY"))

Detail Query

This is an example of a snippet that may be added to something that loads detailed data such as a Jupyter Notebook. Detailed data is essentially granular, non-aggregated data from Visier entities. For example, subjects such as Employee or events such as Compensation_Payout.

with VisierSession(auth) as s:
    client = QueryApiClient(s)
    # List query from JSON query definition
    list_query = load_json("detail/employee-pay_level.json")
    list_result = client.list(list_query)
    df_list = pd.DataFrame.from_records(data=list_result.rows(), columns=list_result.header)

    # ...
    print(df_list.head)

Aggregate Query

Aggregate queries execute queries around Visier's predefined metrics. A metric is a calculation that targets a specific quantifiable question or scenario. They range from very simple like employeeCount to more complex ones like hrRecruitingBudgetedLaborCostPerFTE.

With a VisierSession available, an aggregate query is executed functionally identically:

with VisierSession(auth) as s:
    client = QueryApiClient(s)
    # Aggregate query from JSON query definition
    aggregate_query = load_json("aggregate/applicants-source.json")
    aggregate_result = client.aggregate(aggregate_query)
    df_aggregate = pd.DataFrame.from_records(data=aggregate_result.rows(), columns=aggregate_result.header)

    # Now that the data is in a Pandas Data Frame, do something with it, or just...
    print(df_aggregate.head)

SQL-like Queries

SQL-like allows definition of both aggregate as well as detail queries:

Detail Query

with VisierSession(auth) as s:
    client = QueryApiClient(s)
    # SQL-like detail query
    sql_detail_query = load_str("sql-like/detail/employee-demo.sql")
    list_result = client.sqllike(sql_detail_query)
    df_list = pd.DataFrame.from_records(data=list_result.rows(), columns=list_result.header)

    # ...
    print(df_list.head)

Aggregate Query

This example shows the query definition. Notice how the options object can be used to aggressively eliminate zero and null-valued cells for the purpose of reducing the size of the overall result set to only include rows whose metric value > 0.

with VisierSession(auth) as s:
    client = QueryApiClient(s)
    # SQL-like aggregate query
    sql_aggregate_query = load_str("sql-like/aggregate/employee-count.sql")
    sparse_options = load_json("sql-like/options/sparse.json")
    aggregate_result = client.sqllike(sql_aggregate_query, sparse_options)
    df_aggregate = pd.DataFrame.from_records(data=aggregate_result.rows(), columns=aggregate_result.header)

    # ...
    print(df_aggregate.head)

Model API

The Model API Client is used to make calls to the Visier Model API. In order to run the example below, ensure you add the following import statement to your program:

from visier.api import ModelApiClient

In the example below, we query for the metadata for two named selection concepts on the Requisition analytic object:

    with VisierSession(auth) as session:
        model_client = ModelApiClient(session)

        concepts = model_client.get_selection_concepts("Requisition", ["isRequisitionbyOtherIncomingReasons", "isActiveRequisition"])
        print(concepts)

Direct Intake API

The Direct Intake API enable clients to load data whose structure already matches the target analytic object. Be sure to import the appropriate API client: from visier.api import DirectIntakeApiClient

The instantiation of the API client follows the same pattern as both Query and Model. Regarding the semantics of the API, there are two points to be mindful of:

1. The Direct Intake API is so called because this method of loading data into the Visier system relies on the source data already having been cleansed, deduplicated and transformed. Should these criteria not be met, then these APIs are not suitable for loading data, and alternative methods that leverage Visier's Data Provisioning data transformation mechanisms should be used instead.
2. The call sequence follows a transactional pattern. A transaction is started, followed by a number of uploads after which the transaction is either committed or, in cases where the load should be aborted, rolled-back.

:warning: Please be sure to read the product documentation to ensure the API calling principal has sufficient capabiltities to successfully make these calls.

Schema determination

As this load mechanism is strictly dependent on the structure of the source files matching the schema of the target objects, a schemas API is available to query for the so called 'staging' schema of the target object:

schema = intake_client.get_object_schema("Employee_Exit")

It's important to note that this schema is distinct from the so called 'analytic' schema obtained through the Model API. The 'analytic' schema will include elements that are used during query composition and will include artifacts whose values are derived from others. The 'staging' schema on the other hand, contains only key fields, simple properties, dimension and reference keys.

Load example

Below is a simple example that shows loading data for Employee and Employee_Exit:

    with VisierSession(auth) as session:
        intake_client = DirectIntakeApiClient(session, raise_on_error=True)

        try:
            response = intake_client.start_transaction()
            tx_id = response.json()["transactionId"]
            print(f"Transaction ID: {tx_id}")

            response = intake_client.upload_file(tx_id, "Employee", "/tmp/data/employee-data.zip")
            print(response)

            response = intake_client.upload_file(tx_id, "Employee_Exit", "/tmp/data/exits.csv")
            print(response)
            
            response = intake_client.commit_transaction(tx_id)
            print(response)
        except:
            print(f"Intake failed. Rolling back {tx_id}")
            intake_client.rollback_transaction(tx_id)

Any Visier public API

While connector provides specific functions for querying data, it also provides a lower level, generic function for executing other public Visier APIs. Below is a simple example for determining which Plans have been defined for a given model:

def get_location_levels(context: SessionContext) -> Response:
    path = "/v1/data/model/plan-models/WorkforcePlanModel/plans"
    return context.session().get(url=context.mk_url(path))

with VisierSession(auth) as s:
    levels = s.execute(get_location_levels)
    print(levels.json())

Installation

Add visier-connector as a dependency to your module or install directly: pip install -U visier-connector