Concurrency & Multi-tenancy#

To safely use pygraphistry in concurrent and multitenant settings, use client objects. Use of top-level calls like register() and plot() are unsafe in these settings as they use global variables.

Client objects automatically isolate session state like graphistry server and database tokens. Each client and derived plottable objects are safe for use within a single concurrency context like a thread or event loop, and you can have multiple in different threads and event loops.

Creating Client Objects#

import graphistry

# Create independent client instances
alice_g = graphistry.client()
alice_g.register(api=3, username='alice', password='pw')

bob_g = graphistry.client()
bob_g.register(api=3, username='bob', password='pw')

Multi-tenant Example#

Different users can work with their own isolated sessions:

import graphistry

# Alice's client with her credentials and settings
alice_g = graphistry.client()
alice_g.register(api=3, username='alice', password='alice_pw')
alice_g.privacy(mode='public')

# Bob's client with his credentials and settings
bob_g = graphistry.client()
bob_g.register(api=3, username='bob', password='bob_pw')
bob_g.privacy(mode='org')

# Each client creates isolated plottables
alice_plot = alice_g.edges(alice_data).plot(render=False)
bob_plot = bob_g.edges(bob_data).plot(render=False)

Multiple Servers Example#

Connect to different Graphistry servers (e.g., staging vs production):

import graphistry

# Production server client
prod_g = graphistry.client()
prod_g.register(
    api=3,
    server='prod.graphistry.com',
    username='user',
    password='pw'
)

# Staging server client
staging_g = graphistry.client()
staging_g.register(
    api=3,
    server='staging.graphistry.com',
    username='user',
    password='pw'
)

# Use different servers for different purposes
prod_url = prod_g.edges(production_data).plot(render=False)
staging_url = staging_g.edges(test_data).plot(render=False)

Multi-threaded Example#

Each thread should use its own client instance:

import graphistry
import threading

def process_user_data(username, password, data):
    # Each thread creates its own client
    g = graphistry.client()
    g.register(api=3, username=username, password=password)

    # Process and plot data
    url = g.edges(data).plot(render=False)
    return url

# Launch threads with separate clients
threads = []
for user_info in users:
    t = threading.Thread(
        target=process_user_data,
        args=(user_info['username'], user_info['password'], user_info['data'])
    )
    threads.append(t)
    t.start()

# Wait for all threads to complete
for t in threads:
    t.join()

Transferring Plottables Between Clients#

You can transfer ownership of a plottable from one client to another:

import graphistry

# Create plottable with Alice's client
alice_g = graphistry.client()
alice_g.register(api=3, username='alice', password='alice_pw')
g = alice_g.edges(data)

# Transfer to Bob's client
bob_g = graphistry.client()
bob_g.register(api=3, username='bob', password='bob_pw')
g_bob = bob_g.set_client_for(g)

# Now the plottable uses Bob's credentials and settings
url = g_bob.plot(render=False)

Performance Considerations#

Creating a client requires an authentication round trip, unless you set the JWT token manually
Each plot() call may refresh authentication tokens

Concurrency & Multi-tenancy

Contents