int(retry_after) raises ValueError on a non-integer Retry-After value

Summary

The 429 retry handlers in admin_graphql_request coerce Retry-After with int(), which raises ValueError on any value that is not a bare integer. HTTP permits a decimal seconds value as well as an HTTP-date, and decimal values are what Shopify's own rate-limited REST responses carry.

The exception is uncaught, so it propagates out of admin_graphql_request instead of being retried or returned as a GQLResult.

Version: shopifyapp 1.0.1 (sdist from PyPI).

Cause

shopify_app/graphql/admin_graphql.py:398 (sync):

time.sleep(int(retry_after))

and shopify_app/graphql/admin_graphql.py:454 (async):

await asyncio.sleep(int(retry_after))

Reproduction

>>> int("2.0")
Traceback (most recent call last):
  ...
ValueError: invalid literal for int() with base 10: '2.0'

Currently masked

This is not reachable today, because the header is never actually read: response_headers.get("Retry-After", "1") always returns its "1" default, for the reason described in #17.

That coupling is the reason this is worth filing on its own — repairing the header lookup alone turns a silent bug into an uncaught exception. The two are best addressed together.

Suggested fix

Parse defensively and fall back rather than raising:

try:
    delay = float(retry_after)
except (TypeError, ValueError):
    delay = 1.0
time.sleep(delay)

float() accepts the integer form as well, so it covers both spellings. If HTTP-date support is wanted, email.utils.parsedate_to_datetime handles that form.

Beginne in shopify_app/graphql/admin_graphql.py bei den synchronen und asynchronen Retry-Handlern von admin_graphql_request um die Zeilen 398 und 454. Verfolge, wie Retry-After abgerufen wird, und überprüfe anschließend, dass dezimale oder ungültige Werte nicht mehr als ValueError nach außen gelangen und dass beide Handler ihr Retry- oder GQLResult-Verhalten weiterhin beibehalten.