lumipy 1.0.285

Python library for Luminesce

Lumipy

[loom-ee-pie]

Introduction

Lumipy is a python library that integrates Luminesce and the Python Data Science Stack. It is designed to be used in Jupyter, but you can use it scripts and modules as well.

It has two components

• Getting data: a fluent syntax for scripting up queries using python code. This makes it easy to build complex queries and get your data back as pandas DataFrames.
• Integration: infrastructure to build providers in python. This allows you to build data sources and transforms such as ML models and connect them to Luminesce. They can then be used by other users from the Web UI, Power BI, etc.

Lumipy is designed to be as easy to use and as unobtrusive as possible. You should have to do minimal imports and everything should be explorable from Jupyter through tab completion and shift + tab.

Install

Lumipy is available from PyPI:

LumiPy Is our latest package which utilises the V2 Finbourne SDKs.

It is important to uninstall dve-lumipy-preview before installing lumipy. You can do this by running:

pip uninstall dve-lumipy-preview

We recommend using the --force-reinstall option to make this transition smoother. Please note that this will force update all dependencies for lumipy and could affect your other Python projects.

pip install --force-reinstall lumipy

If you prefer not to update all dependencies, you can omit the --force-reinstall and use the regular pip install command instead:

pip install lumipy

Dve-Lumipy-Preview uses the V1 Finbourne SDKs and is no longer maintained.

pip install dve-lumipy-preview

Configure

Add a personal access token to your config. This first one will be the active one.

import lumipy as lm

lm.config.add('fbn-prd', '<your PAT>')

If you add another domain and PAT you will need to switch to it.

import lumipy as lm

lm.config.add('fbn-ci', '<your PAT>')
lm.config.domain = 'fbn-ci'

Query

All built around the atlas object. This is the starting point for exploring your data sources and then using them.

Build your atlas with lm.get_atlas. If you don't supply credentials it will default to your active domain in the config. If there is no active domain in your config it will fall back to env vars.

import lumipy as lm

atlas = lm.get_atlas()

ins = atlas.lusid_instrument()
ins.select('^').limit(10).go()

You can also specify the domain here by a positional argument, e.g. lm.get_atlas('fbn-ci') will use fbn-ci and will override the active domain.

Client objects are created in the same way. You can submit raw SQL strings as queries using run()

import lumipy as lm

client = lm.get_client()
client.run('select ^ from lusid.instrument limit 10')

You can create a client or atlas for a domain other than the active one by specifying it in get_client or get_atlas.

import lumipy as lm

client = lm.get_client('fbn-prd')
atlas = lm.get_atlas('fbn-prd')

Connect

Python providers are build by inheriting from a base class, BaseProvider, and implementing the __init__ and get_data methods. The former defines the 'shape' of the output data and the parameters it takes. The latter is where the provider actually does something. This can be whatever you want as long as it returns a dataframe with the declared columns.

Running Providers

This will run the required setup on the first startup. Once that's finished it'll spin up a provider that returns Fisher's Irises dataset. Try it out from the web GUI, or from an atlas in another notebook. Remember to get the atlas again once it's finished starting up.

This uses the built-in PandasProvider class to make a provider object, adds it to a ProviderManager and then starts it.

import lumipy.provider as lp

p = lp.PandasProvider(
    'https://raw.githubusercontent.com/mwaskom/seaborn-data/master/iris.csv',
    'iris'
)
lp.ProviderManager(p).run()

This will also default to the active domain if none is specified as an argument to the provider manager.

You can run globally in the domain so other users can query your provider by setting user='global' and whitelist_me=True in the ProviderManager constructor.

The setup consists of getting the dlls for the dotnet app (provider factory) and getting the pem files to run in the domain. To run the setup on its own run the lp.setup() function. This takes the same arguments as get_client and get_atlas.

Building Providers

The following example will simulate a set of coin flips. It has two columns Label and Result, and one parameter Probability with a default value of 0.5.

Its name and column/param content are specified in __init__. The simulation of the coin flips happens inside get_data where we draw numbers from a binomial distribution with the given probability and n = 1.

We also have a check for the probability value. If it's out of range an error will be thrown in python and reported back in the progress log and query status.

Finally, the provider object is instantiated and given to a provider manager. The provider manager is then started up with the run() method.

import lumipy.provider as lp
from pandas import DataFrame
from typing import Union, Iterator
import numpy as np


class CoinFlips(lp.BaseProvider):

    def __init__(self):

        columns = [
            lp.ColumnMeta('Label', lp.DType.Text),
            lp.ColumnMeta('Result', lp.DType.Int),
        ]

        params = [lp.ParamMeta('Probability', lp.DType.Double, default_value=0.5)]

        super().__init__('test.coin.flips', columns, params)

    def get_data(self, context) -> Union[DataFrame, Iterator[DataFrame]]:

        # If no limit is given, default to 100 rows.
        limit = context.limit()
        if limit is None:
            limit = 100

        # Get param value from params dict. If it's out of bounds throw an error. 
        p = context.get('Probability')
        if not 0 <= p <= 1:
            raise ValueError(f'Probability must be between 0 and 1. Was {p}.')

        # Generate the coin flips and return. 
        return DataFrame({'Label':f'Flip {i}', 'Result': np.random.binomial(1, p)} for i in range(limit))


coin_flips = CoinFlips()

lp.ProviderManager(coin_flips).run()

CLI

Lumipy also contains a command line interface (CLI) app with five different functions. You can view help for the CLI and each of the actions using --help. Try this to start with

  $ lumipy --help

Config

This lets you configure your domains and PATs. You can show, add, set, delete and deactivate domains. To see all options and args run the following

  $ lumipy config --help

Config Examples

set Set a domain as the active one.

  $ lumipy config set --domain=my-domain

add Add a domain and PAT to your config.

  $ lumipy config add --domain=my-domain --token=<my token>
  (--overwrite)

show Show a censored view of the config contents.

  $ lumipy config show

delete Delete a domain from the config.

  $ lumipy config delete --domain=my-domain

deactivate Deactivate the config so no domain is used by default.

  $ lumipy config deactivate

Run

This lets you run python providers. You can run prebuilt named sets, CSV files, python files containing provider objects, or even a directory containing CSVs and py files.

Run Examples

.py File

  $ lumipy run path/to/my_providers.py

.csv File

  $ lumipy run path/to/my_data.csv 

Built-in Set

  $ lumipy run demo

Directory

  $ lumipy run path/to/dir

Query

This command runs a SQL query, gets the result back, shows it on screen and then saves it as a CSV.

Query Examples

Run a query (saves as CSV to a temp directory).

  $ lumipy query --sql="select ^ from lusid.instrument limit 5"

Run a query to a defined location.

  $ lumipy query --sql="select ^ from lusid.instrument limit 5" --save-to=/path/to/output.csv

Setup

This lets you run the provider infrastructure setup on your machine.

Setup Examples

Run the py providers setup. This will redownload the certs and get the latest dlls, overwriting any that are already there.

  $ lumipy setup --domain=my-domain

Test

This lets you run the Lumipy test suites.

Test Examples

You can run unit tests, integration tests, provider tests, or everything.

  $ lumipy test unit

Windows Setup

To use LumiPy and run local providers it is recommended that you use an admin powershell terminal.

Install (or update) LumiPy using your powerhsell terminal.

LumiPy (V2 Finbourne SDK)

  $ pip install lumipy --upgrade

Verify that your install was succesful.

  $ lumipy --help

Setup your config with a personal access token (PAT).

  $ lumipy config add --domain=my-domain --token=my-pat-token

Ensure you can run local providers. To run these providers globally add --user==global and --whitelist-me to the command below.

  $ lumipy run demo

Testing Local Changes on Windows

To test your local dve-lumipy changes on Windows add dve-lumipy to your python path (inside your environment variables).

Authenticating with the SDK (Lumipy)

Example using the lumipy.client.get_client() method:

from lumipy.client import get_client
client = get_client()

Recommended Method

Authenticate by setting up the PAT token via the CLI or directly in Python (see the Configure section above).

Secrets File

Initialize get_client using a secrets file:

client = get_client(api_secrets_file="secrets_file_path/secrets.json")

File structure should be:

{
  "api": {
    "luminesceUrl": "https://fbn-ci.lusid.com/honeycomb/",
    "clientId": "clientId",
    "clientSecret": "clientSecret",
    "appName": "appName",
    "certificateFilename": "test_certificate.pem",
    "accessToken": "personal access token"
  },
  "proxy": {
    "address": "http://myproxy.com:8080",
    "username": "proxyuser",
    "password": "proxypass"
  }
}

Keyword Arguments

Initialize get_client with keyword arguments:

client = get_client(username="myusername", ...)

Relevant keyword arguments include:

• token_url
• api_url
• username
• password
• client_id
• client_secret
• app_name
• certificate_filename
• proxy_address
• proxy_username
• proxy_password
• access_token

Environment Variables

The following environment variables can also be set:

• FBN_TOKEN_URL
• FBN_LUMINESCE_API_URL
• FBN_USERNAME
• FBN_PASSWORD
• FBN_CLIENT_ID
• FBN_CLIENT_SECRET
• FBN_APP_NAME
• FBN_CLIENT_CERTIFICATE
• FBN_PROXY_ADDRESS
• FBN_PROXY_USERNAME
• FBN_PROXY_PASSWORD
• FBN_ACCESS_TOKEN