concur

The concur module provides concurrent execution primitives for running functions in parallel.

Functions

Function	Returns	Description
concur.map(items, func, workers=0, timeout="", on_error="abort")	list	Apply func to each item concurrently, returning results in order
concur.each(items, func, workers=0, timeout="", on_error="abort")	None	Apply func to each item concurrently (no return values)
concur.exec(*fns, timeout="", on_error="abort")	tuple	Execute multiple functions concurrently, returning their results

Parameters

Parameter	Default	Description
workers	0	Maximum number of concurrent workers. 0 = unbounded (one goroutine per item)
timeout	""	Maximum time to wait (e.g. "30s", "5m"). Empty = no timeout
on_error	"abort"	Error strategy: "abort" stops on first error; "continue" collects all results
items		Any iterable: list, tuple, or other Starlark iterable

When on_error="continue", failed items return Result objects (with .ok, .value, .error) instead of raising.

Examples

Parallel map

def fetch_status(host):
    result = exec("ping -c1 -W1 " + host)
    return result.ok

hosts = ["host-1", "host-2", "host-3", "host-4"]
statuses = concur.map(hosts, fetch_status, workers=4)
for host, ok in zip(hosts, statuses):
    print(host, "->", "up" if ok else "down")

Parallel each (side effects only)

def deploy(service):
    exec("kubectl rollout restart deployment/" + service)
    log.info("restarted", {"service": service})

services = ["frontend", "backend", "worker"]
concur.each(services, deploy, workers=3, timeout="2m")

Execute independent functions

def get_pods():
    return exec("kubectl get pods -o json").stdout

def get_nodes():
    return exec("kubectl get nodes -o json").stdout

def get_services():
    return exec("kubectl get svc -o json").stdout

pods, nodes, services = concur.exec(get_pods, get_nodes, get_services, timeout="30s")

Error handling with continue

def risky_operation(item):
    if item == "bad":
        fail("item is bad")
    return item.upper()

results = concur.map(
    ["good", "bad", "fine"],
    risky_operation,
    on_error="continue",
)

for r in results:
    if r.ok:
        print("Success:", r.value)
    else:
        print("Failed:", r.error)

Bounded concurrency

# Limit to 2 concurrent API calls
urls = ["https://api.example.com/a", "https://api.example.com/b", "https://api.example.com/c"]
responses = concur.map(urls, http.get, workers=2)

Note: All concur functions that can fail support try_ variants that return a Result instead of raising an error.