jpawlowski/hass.tibber_prices
1
0
Fork 0
mirror of https://github.com/jpawlowski/hass.tibber_prices.git synced 2026-07-27 17:26:48 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
hass.tibber_prices/custom_components/tibber_prices/api/client.py
Julian Pawlowski 8da3083fb2 fix(api): accept priceInfoRange-only responses and retry transient empty data 
Some markets/accounts (e.g. NL hourly) return an empty
priceInfo(QUARTER_HOURLY).today while priceInfoRange still delivers the full
set of intervals. _check_price_info_empty required today's data and wrongly
classified such complete responses as empty, which blocked setup. It now
treats a response as valid when today OR yesterday data is present, since
priceInfoRange is the authoritative source the interval pool uses.

Additionally, "Empty data received" is now retried a few times with
exponential backoff, smoothing brief Tibber outages that transiently return
empty data within a single update cycle. Long outages are still absorbed by
the IntervalPool cache fallback, so the in-request retry count stays small.

Reported in #141.

Impact: Setup and updates no longer fail with "Empty data received" for NL
hourly accounts (and similar), and short Tibber blips are smoothed over.
2026-05-30 14:46:13 +00:00

984 lines
38 KiB
Python
Raw Blame History

"""Tibber API Client."""
from __future__ import annotations
import asyncio
import base64
from datetime import datetime, timedelta
import logging
import re
import socket
from typing import TYPE_CHECKING, Any
from zoneinfo import ZoneInfo
import aiohttp
from homeassistant.util import dt as dt_util
from .exceptions import (
TibberPricesApiClientAuthenticationError,
TibberPricesApiClientCommunicationError,
TibberPricesApiClientError,
TibberPricesApiClientPermissionError,
)
from .helpers import flatten_price_info, prepare_headers, verify_graphql_response, verify_response_or_raise
from .queries import TibberPricesQueryType
if TYPE_CHECKING:
from custom_components.tibber_prices.coordinator.time_service import TibberPricesTimeService
_LOGGER = logging.getLogger(__name__)
_LOGGER_API_DETAILS = logging.getLogger(__name__ + ".details")
class TibberPricesApiClient:
"""Tibber API Client."""
def __init__(
self,
access_token: str,
session: aiohttp.ClientSession,
version: str,
) -> None:
"""Tibber API Client."""
self._access_token = access_token
self._session = session
self._version = version
self._request_semaphore = asyncio.Semaphore(2) # Max 2 concurrent requests
self.time: TibberPricesTimeService | None = None # Set externally by coordinator (optional during config flow)
self._last_request_time = None # Set on first request
self._min_request_interval = timedelta(seconds=1) # Min 1 second between requests
self._max_retries = 5
self._retry_delay = 2 # Base retry delay in seconds
# Empty-data responses are usually permanent (no data for the requested
# range), but during a Tibber outage the API can return empty data
# transiently before recovering. Retry a small number of times with backoff
# so a brief blip is smoothed out within a single update cycle.
#
# NOTE: This is NOT the mechanism that bridges multi-hour outages. Long
# outages are absorbed by the IntervalPool cache fallback (see
# interval_pool/manager.py): as long as cached intervals cover the current
# interval, the integration keeps serving cached prices and retries on the
# NEXT update cycle. That bridge is unbounded in time, so this in-request
# retry count must stay small to avoid blocking the event loop.
self._max_empty_data_retries = 2
# Timeout configuration - more granular control
self._connect_timeout = 10 # Connection timeout in seconds
self._request_timeout = 25 # Total request timeout in seconds
self._socket_connect_timeout = 5 # Socket connection timeout
async def async_get_viewer_details(self) -> Any:
"""Get comprehensive viewer and home details from Tibber API."""
return await self._api_wrapper(
data={
"query": """
{
viewer {
userId
name
login
accountType
homes {
id
type
appNickname
appAvatar
size
timeZone
mainFuseSize
numberOfResidents
primaryHeatingSource
hasVentilationSystem
address {
address1
address2
address3
postalCode
city
country
latitude
longitude
}
owner {
id
firstName
lastName
isCompany
name
contactInfo {
email
mobile
}
language
}
meteringPointData {
consumptionEan
gridCompany
gridAreaCode
priceAreaCode
productionEan
energyTaxType
vatType
estimatedAnnualConsumption
}
currentSubscription {
id
status
validFrom
validTo
priceInfo {
current {
currency
}
}
}
features {
realTimeConsumptionEnabled
}
}
}
}
"""
},
query_type=TibberPricesQueryType.USER,
)
async def async_get_price_info_for_range(
self,
home_id: str,
user_data: dict[str, Any],
start_time: datetime,
end_time: datetime,
) -> dict:
"""
Get price info for a specific time range with automatic routing.
This is a convenience wrapper around interval_pool.get_price_intervals_for_range().
Args:
home_id: Home ID to fetch price data for.
user_data: User data dict containing home metadata (including timezone).
start_time: Start of the range (inclusive, timezone-aware).
end_time: End of the range (exclusive, timezone-aware).
Returns:
Dict with "home_id" and "price_info" (list of intervals).
Raises:
TibberPricesApiClientError: If arguments invalid or requests fail.
"""
# Import here to avoid circular dependency (interval_pool imports TibberPricesApiClient)
from custom_components.tibber_prices.interval_pool import get_price_intervals_for_range # noqa: PLC0415
price_info = await get_price_intervals_for_range(
api_client=self,
home_id=home_id,
user_data=user_data,
start_time=start_time,
end_time=end_time,
)
return {
"home_id": home_id,
"price_info": price_info,
}
async def async_get_price_info(self, home_id: str, user_data: dict[str, Any]) -> dict:
"""
Get price info for a single home.
Uses timezone-aware cursor calculation based on the home's actual timezone
from Tibber API (not HA system timezone). This ensures correct "day before yesterday
midnight" calculation for homes in different timezones.
Args:
home_id: Home ID to fetch price data for.
user_data: User data dict containing home metadata (including timezone).
REQUIRED - must be fetched before calling this method.
Returns:
Dict with "home_id", "price_info", and other home data.
Raises:
TibberPricesApiClientError: If TimeService not initialized or user_data missing.
"""
if not self.time:
msg = "TimeService not initialized - required for price info processing"
raise TibberPricesApiClientError(msg)
if not user_data:
msg = "User data required for timezone-aware price fetching - fetch user data first"
raise TibberPricesApiClientError(msg)
if not home_id:
msg = "Home ID is required"
raise TibberPricesApiClientError(msg)
# Build home_id -> timezone mapping from user_data
home_timezones = self._extract_home_timezones(user_data)
# Get timezone for this home (fallback to HA system timezone)
home_tz = home_timezones.get(home_id)
# Calculate cursor: day before yesterday midnight in home's timezone
cursor = self._calculate_cursor_for_home(home_tz)
# Simple single-home query (no alias needed)
query = f"""
{{viewer{{
home(id: "{home_id}") {{
id
currentSubscription {{
priceInfoRange(resolution:QUARTER_HOURLY, first:192, after: "{cursor}") {{
pageInfo{{ count }}
edges{{node{{
startsAt total energy tax level
}}}}
}}
priceInfo(resolution:QUARTER_HOURLY) {{
today{{startsAt total energy tax level}}
tomorrow{{startsAt total energy tax level}}
}}
}}
}}
}}}}
"""
_LOGGER.debug("Fetching price info for home %s", home_id)
data = await self._api_wrapper(
data={"query": query},
query_type=TibberPricesQueryType.PRICE_INFO,
)
# Parse response
viewer = data.get("viewer", {})
home = viewer.get("home")
if not home:
msg = f"Home {home_id} not found in API response"
_LOGGER.warning(msg)
return {"home_id": home_id, "price_info": []}
if "currentSubscription" in home and home["currentSubscription"] is not None:
price_info = flatten_price_info(home["currentSubscription"])
else:
_LOGGER.warning(
"Home %s has no active subscription - price data will be unavailable",
home_id,
)
price_info = []
return {
"home_id": home_id,
"price_info": price_info,
}
async def async_get_price_info_range(
self,
home_id: str,
user_data: dict[str, Any],
start_time: datetime,
end_time: datetime,
) -> dict:
"""
Get historical price info for a specific time range using priceInfoRange endpoint.
Uses the priceInfoRange GraphQL endpoint for flexible historical data queries.
Intended for intervals BEFORE "day before yesterday midnight" (outside PRICE_INFO scope).
Automatically handles API pagination if Tibber limits batch size.
Args:
home_id: Home ID to fetch price data for.
user_data: User data dict containing home metadata (including timezone).
start_time: Start of the range (inclusive, timezone-aware).
end_time: End of the range (exclusive, timezone-aware).
Returns:
Dict with "home_id" and "price_info" (list of intervals).
Raises:
TibberPricesApiClientError: If arguments invalid or request fails.
"""
if not user_data:
msg = "User data required for timezone-aware price fetching - fetch user data first"
raise TibberPricesApiClientError(msg)
if not home_id:
msg = "Home ID is required"
raise TibberPricesApiClientError(msg)
if start_time >= end_time:
msg = f"Invalid time range: start_time ({start_time}) must be before end_time ({end_time})"
raise TibberPricesApiClientError(msg)
_LOGGER_API_DETAILS.debug(
"fetch_price_info_range called with: start_time=%s (type=%s, tzinfo=%s), end_time=%s (type=%s, tzinfo=%s)",
start_time,
type(start_time),
start_time.tzinfo,
end_time,
type(end_time),
end_time.tzinfo,
)
# Calculate cursor and interval count
start_cursor = self._encode_cursor(start_time)
interval_count = self._calculate_interval_count(start_time, end_time)
_LOGGER_API_DETAILS.debug(
"Calculated cursor for range: start_time=%s, cursor_time=%s, encoded=%s",
start_time,
start_time,
start_cursor,
)
# Fetch all intervals with automatic paging
price_info = await self._fetch_price_info_with_paging(
home_id=home_id,
start_cursor=start_cursor,
interval_count=interval_count,
)
return {
"home_id": home_id,
"price_info": price_info,
}
def _calculate_interval_count(self, start_time: datetime, end_time: datetime) -> int:
"""Calculate number of intervals needed based on date range."""
time_diff = end_time - start_time
resolution_change_date = datetime(2025, 10, 1, tzinfo=start_time.tzinfo)
if start_time < resolution_change_date:
# Pre-resolution-change: hourly intervals only
interval_count = int(time_diff.total_seconds() / 3600) # 3600s = 1h
_LOGGER_API_DETAILS.debug(
"Time range is pre-2025-10-01: expecting hourly intervals (count: %d)",
interval_count,
)
else:
# Post-resolution-change: quarter-hourly intervals
interval_count = int(time_diff.total_seconds() / 900) # 900s = 15min
_LOGGER_API_DETAILS.debug(
"Time range is post-2025-10-01: expecting quarter-hourly intervals (count: %d)",
interval_count,
)
return interval_count
async def _fetch_price_info_with_paging(
self,
home_id: str,
start_cursor: str,
interval_count: int,
) -> list[dict[str, Any]]:
"""
Fetch price info with automatic pagination if API limits batch size.
GraphQL Cursor Pagination:
- endCursor points to the last returned element (inclusive)
- Use 'after: endCursor' to get elements AFTER that cursor
- If count < requested, more pages available
Args:
home_id: Home ID to fetch price data for.
start_cursor: Base64-encoded start cursor.
interval_count: Total number of intervals to fetch.
Returns:
List of all price interval dicts across all pages.
"""
price_info = []
remaining_intervals = interval_count
cursor = start_cursor
page = 0
while remaining_intervals > 0:
page += 1
# Fetch one page
page_data = await self._fetch_single_page(
home_id=home_id,
cursor=cursor,
requested_count=remaining_intervals,
page=page,
)
if not page_data:
break
# Extract intervals and pagination info
page_intervals = page_data["intervals"]
returned_count = page_data["count"]
end_cursor = page_data["end_cursor"]
has_next_page = page_data.get("has_next_page", False)
price_info.extend(page_intervals)
_LOGGER_API_DETAILS.debug(
"Page %d: Received %d intervals for home %s (total so far: %d/%d, endCursor=%s, hasNextPage=%s)",
page,
returned_count,
home_id,
len(price_info),
interval_count,
end_cursor,
has_next_page,
)
# Update remaining count
remaining_intervals -= returned_count
# Check if we need more pages
# Continue if: (1) we still need more intervals AND (2) API has more data
if remaining_intervals > 0 and end_cursor:
cursor = end_cursor
_LOGGER_API_DETAILS.debug(
"Still need %d more intervals - fetching next page with cursor %s",
remaining_intervals,
cursor,
)
else:
# Done: Either we have all intervals we need, or API has no more data
if remaining_intervals > 0:
_LOGGER.warning(
"API has no more data - received %d out of %d requested intervals (missing %d)",
len(price_info),
interval_count,
remaining_intervals,
)
else:
_LOGGER.debug(
"Pagination complete - received all %d requested intervals",
interval_count,
)
break
_LOGGER_API_DETAILS.debug(
"Fetched %d total historical intervals for home %s across %d page(s)",
len(price_info),
home_id,
page,
)
return price_info
async def _fetch_single_page(
self,
home_id: str,
cursor: str,
requested_count: int,
page: int,
) -> dict[str, Any] | None:
"""
Fetch a single page of price intervals.
Args:
home_id: Home ID to fetch price data for.
cursor: Base64-encoded cursor for this page.
requested_count: Number of intervals to request.
page: Page number (for logging).
Returns:
Dict with "intervals", "count", and "end_cursor" keys, or None if no data.
"""
query = f"""
{{viewer{{
home(id: "{home_id}") {{
id
currentSubscription {{
priceInfoRange(resolution:QUARTER_HOURLY, first:{requested_count}, after: "{cursor}") {{
pageInfo{{
count
hasNextPage
startCursor
endCursor
}}
edges{{
cursor
node{{
startsAt total energy tax level
}}
}}
}}
}}
}}
}}}}
"""
_LOGGER_API_DETAILS.debug(
"Fetching historical price info for home %s (page %d): %d intervals from cursor %s",
home_id,
page,
requested_count,
cursor,
)
data = await self._api_wrapper(
data={"query": query},
query_type=TibberPricesQueryType.PRICE_INFO_RANGE,
)
# Parse response
viewer = data.get("viewer", {})
home = viewer.get("home")
if not home:
_LOGGER.warning("Home %s not found in API response", home_id)
return None
if "currentSubscription" not in home or home["currentSubscription"] is None:
_LOGGER.warning("Home %s has no active subscription - price data will be unavailable", home_id)
return None
# Extract priceInfoRange data
subscription = home["currentSubscription"]
price_info_range = subscription.get("priceInfoRange", {})
page_info = price_info_range.get("pageInfo", {})
edges = price_info_range.get("edges", [])
# Flatten edges to interval list
intervals = [edge["node"] for edge in edges if "node" in edge]
return {
"intervals": intervals,
"count": page_info.get("count", len(intervals)),
"end_cursor": page_info.get("endCursor"),
"has_next_page": page_info.get("hasNextPage", False),
}
def _extract_home_timezones(self, user_data: dict[str, Any]) -> dict[str, str]:
"""
Extract home_id -> timezone mapping from user_data.
Args:
user_data: User data dict from async_get_viewer_details() (required).
Returns:
Dict mapping home_id to timezone string (e.g., "Europe/Oslo").
"""
home_timezones = {}
viewer = user_data.get("viewer", {})
homes = viewer.get("homes", [])
for home in homes:
home_id = home.get("id")
timezone = home.get("timeZone")
if home_id and timezone:
home_timezones[home_id] = timezone
_LOGGER_API_DETAILS.debug("Extracted timezone %s for home %s", timezone, home_id)
elif home_id:
_LOGGER.warning("Home %s has no timezone in user data, will use fallback", home_id)
return home_timezones
def _calculate_day_before_yesterday_midnight(self, home_timezone: str | None) -> datetime:
"""
Calculate day before yesterday midnight in home's timezone.
CRITICAL: Uses REAL TIME (dt_util.now()), NOT TimeService.now().
This ensures API boundary calculations are based on actual current time,
not simulated time from TimeService.
Args:
home_timezone: Timezone string (e.g., "Europe/Oslo").
If None, falls back to HA system timezone.
Returns:
Timezone-aware datetime for day before yesterday midnight.
"""
# Get current REAL time (not TimeService)
now = dt_util.now()
# Convert to home's timezone or fallback to HA system timezone
if home_timezone:
try:
tz = ZoneInfo(home_timezone)
now_in_home_tz = now.astimezone(tz)
except (KeyError, ValueError, OSError) as error:
_LOGGER.warning(
"Invalid timezone %s (%s), falling back to HA system timezone",
home_timezone,
error,
)
now_in_home_tz = dt_util.as_local(now)
else:
# Fallback to HA system timezone
now_in_home_tz = dt_util.as_local(now)
# Calculate day before yesterday midnight
return (now_in_home_tz - timedelta(days=2)).replace(hour=0, minute=0, second=0, microsecond=0)
def _encode_cursor(self, timestamp: datetime) -> str:
"""
Encode a timestamp as base64 cursor for GraphQL API.
Args:
timestamp: Timezone-aware datetime to encode.
Returns:
Base64-encoded ISO timestamp string.
"""
iso_string = timestamp.isoformat()
return base64.b64encode(iso_string.encode()).decode()
def _parse_timestamp(self, timestamp_str: str) -> datetime:
"""
Parse ISO timestamp string to timezone-aware datetime.
Args:
timestamp_str: ISO format timestamp string.
Returns:
Timezone-aware datetime object.
"""
return dt_util.parse_datetime(timestamp_str) or dt_util.now()
def _calculate_cursor_for_home(self, home_timezone: str | None) -> str:
"""
Calculate cursor (day before yesterday midnight) for a home's timezone.
Convenience wrapper around _calculate_day_before_yesterday_midnight()
and _encode_cursor() for backward compatibility with existing code.
Args:
home_timezone: Timezone string (e.g., "Europe/Oslo", "America/New_York").
If None, falls back to HA system timezone.
Returns:
Base64-encoded ISO timestamp string for use as GraphQL cursor.
"""
day_before_yesterday_midnight = self._calculate_day_before_yesterday_midnight(home_timezone)
return self._encode_cursor(day_before_yesterday_midnight)
async def _make_request(
self,
headers: dict[str, str],
data: dict,
query_type: TibberPricesQueryType,
) -> dict[str, Any]:
"""Make an API request with comprehensive error handling for network issues."""
_LOGGER_API_DETAILS.debug("Making API request with data: %s", data)
try:
# More granular timeout configuration for better network failure handling
timeout = aiohttp.ClientTimeout(
total=self._request_timeout, # Total request timeout: 25s
connect=self._connect_timeout, # Connection timeout: 10s
sock_connect=self._socket_connect_timeout, # Socket connection: 5s
)
response = await self._session.request(
method="POST",
url="https://api.tibber.com/v1-beta/gql",
headers=headers,
json=data,
timeout=timeout,
)
verify_response_or_raise(response)
response_json = await response.json()
_LOGGER_API_DETAILS.debug("Received API response: %s", response_json)
await verify_graphql_response(response_json, query_type)
return response_json["data"]
except aiohttp.ClientResponseError as error:
_LOGGER.exception("HTTP error during API request")
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=str(error))
) from error
except aiohttp.ClientConnectorError as error:
_LOGGER.exception("Connection error - server unreachable or network down")
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=str(error))
) from error
except aiohttp.ServerDisconnectedError as error:
_LOGGER.exception("Server disconnected during request")
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=str(error))
) from error
except TimeoutError as error:
_LOGGER.exception(
"Request timeout after %d seconds - slow network or server overload",
self._request_timeout,
)
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.TIMEOUT_ERROR.format(exception=str(error))
) from error
except socket.gaierror as error:
self._handle_dns_error(error)
raise # Ensure type checker knows this path always raises
except OSError as error:
self._handle_network_error(error)
raise # Ensure type checker knows this path always raises
def _handle_dns_error(self, error: socket.gaierror) -> None:
"""Handle DNS resolution errors with IPv4/IPv6 dual stack considerations."""
error_msg = str(error)
if "Name or service not known" in error_msg:
_LOGGER.exception("DNS resolution failed - domain name not found")
elif "Temporary failure in name resolution" in error_msg:
_LOGGER.exception("DNS resolution temporarily failed - network or DNS server issue")
elif "Address family for hostname not supported" in error_msg:
_LOGGER.exception("DNS resolution failed - IPv4/IPv6 address family not supported")
elif "No address associated with hostname" in error_msg:
_LOGGER.exception("DNS resolution failed - no IPv4/IPv6 addresses found")
else:
_LOGGER.exception("DNS resolution failed - check internet connection: %s", error_msg)
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=str(error))
) from error
def _handle_network_error(self, error: OSError) -> None:
"""Handle network-level errors with IPv4/IPv6 dual stack considerations."""
error_msg = str(error)
errno = getattr(error, "errno", None)
# Common IPv4/IPv6 dual stack network error codes
errno_network_unreachable = 101 # ENETUNREACH
errno_host_unreachable = 113 # EHOSTUNREACH
errno_connection_refused = 111 # ECONNREFUSED
errno_connection_timeout = 110 # ETIMEDOUT
if errno == errno_network_unreachable:
_LOGGER.exception("Network unreachable - check internet connection or IPv4/IPv6 routing")
elif errno == errno_host_unreachable:
_LOGGER.exception("Host unreachable - routing issue or IPv4/IPv6 connectivity problem")
elif errno == errno_connection_refused:
_LOGGER.exception("Connection refused - server not accepting connections")
elif errno == errno_connection_timeout:
_LOGGER.exception("Connection timed out - network latency or server overload")
elif "Address family not supported" in error_msg:
_LOGGER.exception("Address family not supported - IPv4/IPv6 configuration issue")
elif "Protocol not available" in error_msg:
_LOGGER.exception("Protocol not available - IPv4/IPv6 stack configuration issue")
elif "Network is down" in error_msg:
_LOGGER.exception("Network interface is down - check network adapter")
elif "Permission denied" in error_msg:
_LOGGER.exception("Network permission denied - firewall or security restriction")
else:
_LOGGER.exception("Network error - internet may be down: %s", error_msg)
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=str(error))
) from error
async def _handle_request(
self,
headers: dict[str, str],
data: dict,
query_type: TibberPricesQueryType,
) -> Any:
"""Handle a single API request with rate limiting."""
async with self._request_semaphore:
# Rate limiting: ensure minimum interval between requests
if self.time and self._last_request_time:
now = self.time.now()
time_since_last_request = now - self._last_request_time
if time_since_last_request < self._min_request_interval:
sleep_time = (self._min_request_interval - time_since_last_request).total_seconds()
_LOGGER_API_DETAILS.debug(
"Rate limiting: waiting %s seconds before next request",
sleep_time,
)
await asyncio.sleep(sleep_time)
if self.time:
self._last_request_time = self.time.now()
return await self._make_request(
headers,
data or {},
query_type,
)
def _should_retry_error(self, error: Exception, retry: int) -> tuple[bool, int]:
"""Determine if an error should be retried and calculate delay."""
# Check if we've exceeded max retries first
if retry >= self._max_retries:
return False, 0
# Non-retryable errors - authentication and permission issues
if isinstance(
error,
(
TibberPricesApiClientAuthenticationError,
TibberPricesApiClientPermissionError,
),
):
return False, 0
# Handle API-specific errors
if isinstance(error, TibberPricesApiClientError):
return self._handle_api_error_retry(error, retry)
# Network and timeout errors - retryable with exponential backoff
if isinstance(error, (aiohttp.ClientError, socket.gaierror, TimeoutError)):
delay = min(self._retry_delay * (2**retry), 30) # Cap at 30 seconds
return True, delay
# Unknown errors - not retryable
return False, 0
def _handle_api_error_retry(self, error: TibberPricesApiClientError, retry: int) -> tuple[bool, int]:
"""Handle retry logic for API-specific errors."""
error_msg = str(error)
# Non-retryable: Invalid queries and bad requests are permanent client errors.
if "Invalid GraphQL query" in error_msg or "Bad request" in error_msg:
return False, 0
# Empty data is usually permanent (API has no data for the requested range),
# but can be returned transiently during a Tibber outage. Retry a limited
# number of times with exponential backoff before giving up.
if "Empty data received" in error_msg:
if retry < self._max_empty_data_retries:
delay = min(self._retry_delay * (2**retry), 30)
return True, delay
return False, 0
# Rate limits - only retry if server explicitly says so
if "Rate limit exceeded" in error_msg or "rate limited" in error_msg.lower():
delay = self._extract_retry_delay(error, retry)
return True, delay
# Other API errors - not retryable (assume permanent issue)
return False, 0
def _extract_retry_delay(self, error: Exception, retry: int) -> int:
"""Extract retry delay from rate limit error or use exponential backoff."""
error_msg = str(error)
# Try to extract Retry-After value from error message
retry_after_match = re.search(r"retry after (\d+) seconds", error_msg.lower())
if retry_after_match:
try:
retry_after = int(retry_after_match.group(1))
return min(retry_after + 1, 300) # Add buffer, max 5 minutes
except ValueError:
pass
# Try to extract generic seconds value
seconds_match = re.search(r"(\d+) seconds", error_msg)
if seconds_match:
try:
seconds = int(seconds_match.group(1))
return min(seconds + 1, 300) # Add buffer, max 5 minutes
except ValueError:
pass
# Fall back to exponential backoff with cap
base_delay = self._retry_delay * (2**retry)
return min(base_delay, 120) # Cap at 2 minutes for rate limits
async def _api_wrapper(
self,
data: dict | None = None,
headers: dict | None = None,
query_type: TibberPricesQueryType = TibberPricesQueryType.USER,
) -> Any:
"""
Get information from the API with rate limiting and retry logic.
Exception Handling Strategy:
- AuthenticationError: Immediate raise, triggers reauth flow
- PermissionError: Immediate raise, non-retryable
- CommunicationError: Retry with exponential backoff
- ApiClientError (Rate Limit): Retry with Retry-After delay
- ApiClientError (Other): Retry only if explicitly retryable
- Network errors (aiohttp.ClientError, socket.gaierror, TimeoutError):
Converted to CommunicationError and retried
Retry Logic:
- Max retries: 5 (configurable via _max_retries)
- Base delay: 2 seconds (exponential backoff: 2s, 4s, 8s, 16s, 32s)
- Rate limit delay: Uses Retry-After header or falls back to exponential
- Caps: 30s for network errors, 120s for rate limits, 300s for Retry-After
"""
headers = headers or prepare_headers(self._access_token, self._version)
last_error: Exception | None = None
for retry in range(self._max_retries + 1):
try:
return await self._handle_request(headers, data or {}, query_type)
except (
TibberPricesApiClientAuthenticationError,
TibberPricesApiClientPermissionError,
):
_LOGGER.exception("Non-retryable error occurred")
raise
except (
TibberPricesApiClientError,
aiohttp.ClientError,
socket.gaierror,
TimeoutError,
) as error:
last_error = (
error
if isinstance(error, TibberPricesApiClientError)
else TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=str(error))
)
)
should_retry, delay = self._should_retry_error(error, retry)
if should_retry:
error_type = self._get_error_type(error)
_LOGGER.warning(
"Tibber %s error, attempt %d/%d. Retrying in %d seconds: %s",
error_type,
retry + 1,
self._max_retries,
delay,
str(error),
)
await asyncio.sleep(delay)
continue
if "Invalid GraphQL query" in str(error):
_LOGGER.exception("Invalid query - not retrying")
raise
# Handle final error state
if isinstance(last_error, TimeoutError):
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.TIMEOUT_ERROR.format(exception=last_error)
) from last_error
if isinstance(last_error, (aiohttp.ClientError, socket.gaierror)):
raise TibberPricesApiClientCommunicationError(
TibberPricesApiClientCommunicationError.CONNECTION_ERROR.format(exception=last_error)
) from last_error
raise last_error or TibberPricesApiClientError(TibberPricesApiClientError.UNKNOWN_ERROR)
def _get_error_type(self, error: Exception) -> str:
"""Get a descriptive error type for logging."""
if "Rate limit" in str(error):
return "rate limit"
if isinstance(error, (aiohttp.ClientError, socket.gaierror, TimeoutError)):
return "network"
return "API"
Reference in a new issue View git blame Copy permalink