Add QCEW CSV module, ECEC breakdown, response caching + typed errors

Item 1 — QCEW county/industry detail:
- new bls_client/qcew.py: thin client over the QCEW Open Data CSV service
  (area/industry/size slices; no API key, no quota), with filter_rows helper
- covers the county/industry detail the timeseries API can't reach

Item 2 — ECEC benefit breakdown:
- new series.ecec() builder from the authoritative cm.estimate decode table
- real helpers: wages, total benefits, paid leave, supplemental pay,
  health insurance, retirement & savings, legally required + ecec_dashboard()
- FIX: ecec_total_benefits was CMU1036... (education/health industries only);
  correct all-civilian total benefits is CMU1030000000000D

Item 3 — caching + typed errors:
- bls_client/cache.py FileCache (on-disk, TTL); BLSClient(cache=True),
  queries_used counter (cache hits don't spend quota)
- bls_client/errors.py: BLSError / BLSQuotaError / BLSRequestError; quota
  exhaustion now raises a clear BLSQuotaError instead of a generic RuntimeError

Tests: 119 offline (+24) incl. ECEC goldens, QCEW fixture parse, cache/quota
behavior; live smoke extended to ECEC + QCEW. Helper sweep: 96/96 live.
README: caching/QCEW usage, updated coverage + limitations.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-06-22 10:38:12 -04:00
parent 881d62cf32
commit 9e2c55c568
14 changed files with 499 additions and 44 deletions

View File

@ -1,2 +1,3 @@
from .client import BLSClient
from . import queries, series
from .errors import BLSError, BLSRequestError, BLSQuotaError
from . import queries, series, qcew

49
bls_client/cache.py Normal file
View File

@ -0,0 +1,49 @@
"""A tiny on-disk response cache for BLS requests.
BLS data updates monthly/quarterly, so caching identical requests avoids
re-spending the 500/day quota on reruns (reports, the explorer, test sweeps).
Pure stdlib — no extra dependency.
"""
import hashlib
import json
import os
import time
from pathlib import Path
class FileCache:
"""JSON file cache keyed by request payload, with a TTL (seconds)."""
def __init__(self, cache_dir=None, ttl=86400):
self.dir = Path(cache_dir or os.path.expanduser("~/.cache/bls"))
self.ttl = ttl
self.dir.mkdir(parents=True, exist_ok=True)
@staticmethod
def key(payload: dict) -> str:
"""Deterministic key from a request payload (API key excluded, series sorted)."""
relevant = {k: v for k, v in payload.items() if k != "registrationkey"}
if isinstance(relevant.get("seriesid"), list):
relevant["seriesid"] = sorted(relevant["seriesid"])
blob = json.dumps(relevant, sort_keys=True, default=str)
return hashlib.sha256(blob.encode()).hexdigest()
def _path(self, key: str) -> Path:
return self.dir / f"{key}.json"
def get(self, key: str):
p = self._path(key)
if not p.exists():
return None
if self.ttl is not None and (time.time() - p.stat().st_mtime) > self.ttl:
return None
try:
return json.loads(p.read_text())
except (OSError, ValueError):
return None
def set(self, key: str, value) -> None:
try:
self._path(key).write_text(json.dumps(value))
except OSError:
pass # caching is best-effort; never fail a fetch over it

View File

@ -10,14 +10,32 @@ Usage:
import requests
from datetime import datetime
from .cache import FileCache
from .errors import BLSError, BLSRequestError, BLSQuotaError
class BLSClient:
BASE_URL = "https://api.bls.gov/publicAPI/v2"
MAX_SERIES_PER_CALL = 50
MAX_YEARS_PER_CALL = 20
def __init__(self, api_key: str):
def __init__(self, api_key: str, cache=False, cache_dir=None, cache_ttl=86400):
"""
Args:
api_key: BLS registration key.
cache: True to enable the default on-disk cache (~/.cache/bls),
or pass a custom cache object with get(key)/set(key, value).
cache_dir: Override the default cache directory.
cache_ttl: Cache entry lifetime in seconds (default 1 day).
"""
self.api_key = api_key
if cache is True:
self.cache = FileCache(cache_dir, cache_ttl)
elif cache:
self.cache = cache # caller-supplied cache object
else:
self.cache = None
self.queries_used = 0 # network POSTs spent this session (cache hits don't count)
# ------------------------------------------------------------------
# Core fetch
@ -64,23 +82,44 @@ class BLSClient:
"calculations": calculations,
"annualaverage": annual_average,
}
r = requests.post(
f"{self.BASE_URL}/timeseries/data/",
json=payload,
timeout=30,
)
r.raise_for_status()
body = r.json()
if body["status"] != "REQUEST_SUCCEEDED":
raise RuntimeError(f"BLS API error: {body['message']}")
for s in body["Results"]["series"]:
results[s["seriesID"]] = {
"data": s["data"],
"catalog": s.get("catalog", {}),
}
cache_key = self.cache.key(payload) if self.cache else None
batch_result = self.cache.get(cache_key) if cache_key else None
if batch_result is None:
batch_result = self._fetch_batch(payload)
if cache_key:
self.cache.set(cache_key, batch_result)
results.update(batch_result)
return results
def _fetch_batch(self, payload: dict) -> dict:
"""Execute one (uncached) API call and return {series_id: {data, catalog}}."""
r = requests.post(f"{self.BASE_URL}/timeseries/data/", json=payload, timeout=30)
r.raise_for_status()
self.queries_used += 1
body = r.json()
status = body.get("status")
if status != "REQUEST_SUCCEEDED":
msgs = body.get("message", [])
text = " ".join(msgs) if isinstance(msgs, list) else str(msgs)
low = text.lower()
if status == "REQUEST_NOT_PROCESSED" and ("threshold" in low or "exceed" in low):
raise BLSQuotaError(
f"BLS daily query limit reached: {text}", status=status, messages=msgs
)
raise BLSRequestError(
f"BLS API request failed ({status}): {text}", status=status, messages=msgs
)
return {
s["seriesID"]: {"data": s["data"], "catalog": s.get("catalog", {})}
for s in body["Results"]["series"]
}
def fetch_latest(
self,
series_ids: list[str] | str,

23
bls_client/errors.py Normal file
View File

@ -0,0 +1,23 @@
"""Typed exceptions for the BLS client.
Lets callers distinguish "you're throttled" from "your request was bad" — the
two failure modes the old generic RuntimeError conflated.
"""
class BLSError(RuntimeError):
"""Base class for all BLS API errors."""
def __init__(self, message, *, status=None, messages=None):
super().__init__(message)
self.status = status # the API "status" field, e.g. REQUEST_FAILED
self.messages = messages or [] # the API "message" list, verbatim
class BLSRequestError(BLSError):
"""The API rejected the request (malformed series ID, bad year range, etc.)."""
class BLSQuotaError(BLSError):
"""The daily query threshold was exceeded (500/day for a registered key,
25/day unregistered). Resets at midnight Eastern."""

74
bls_client/qcew.py Normal file
View File

@ -0,0 +1,74 @@
"""QCEW Open Data Access — county/industry employment & wages.
The BLS *timeseries* API only carries QCEW national/state totals. The full
QCEW detail (every county, every NAICS industry, establishment counts and
wages) lives behind a separate CSV service that needs no API key and counts
against no quota:
https://data.bls.gov/cew/data/api/{year}/{qtr}/{slice}/{code}.csv
This module is a thin wrapper over that service. It is deliberately separate
from BLSClient (different API, CSV not JSON, no key). Each call returns a list
of plain dict rows (the CSV columns), so it composes with csv/pandas directly.
Quarter argument: 1-4 for a specific quarter, or "a" for annual averages.
Area FIPS examples: "US000" national, "11000" DC (state), "11001" a county,
metro codes start with "C". Industry codes are NAICS (e.g. "10" = all
industries, "722" = food services, "622" = hospitals).
"""
import csv
import io
import requests
QCEW_BASE = "https://data.bls.gov/cew/data/api"
_HEADERS = {"User-Agent": "bls-data-library (https://gitea.doesworks.net/giteaadmin/bls-data)"}
# A few common columns, for reference / convenience:
# area_fips, own_code, industry_code, agglvl_code, size_code, year, qtr,
# annual_avg_estabs, annual_avg_emplvl, total_annual_wages, annual_avg_wkly_wage,
# avg_annual_pay (quarterly slices use month1_emplvl/qtrly_estabs/total_qtrly_wages, etc.)
OWNERSHIP = {
"total": "0", "federal": "1", "state": "2", "local": "3",
"private": "5", "government": "8",
}
def _fetch_csv(url: str) -> list[dict]:
r = requests.get(url, timeout=60, headers=_HEADERS)
r.raise_for_status()
return parse_csv(r.text)
def parse_csv(text: str) -> list[dict]:
"""Parse QCEW CSV text into a list of row dicts (BLS quotes every field)."""
return list(csv.DictReader(io.StringIO(text)))
def area(area_fips: str, year: int, qtr="a") -> list[dict]:
"""Every industry × ownership for one area (county/state/metro/national)."""
return _fetch_csv(f"{QCEW_BASE}/{year}/{qtr}/area/{area_fips}.csv")
def industry(industry_code: str, year: int, qtr="a") -> list[dict]:
"""One NAICS industry across all areas."""
return _fetch_csv(f"{QCEW_BASE}/{year}/{qtr}/industry/{industry_code}.csv")
def size(size_code: str, year: int) -> list[dict]:
"""Establishment-size breakdown (size data is published for Q1 only)."""
return _fetch_csv(f"{QCEW_BASE}/{year}/1/size/{size_code}.csv")
def filter_rows(rows, own_code=None, industry_code=None, agglvl_code=None) -> list[dict]:
"""Filter QCEW rows by ownership / industry / aggregation-level code."""
out = rows
if own_code is not None:
own = OWNERSHIP.get(own_code, own_code)
out = [r for r in out if r.get("own_code") == own]
if industry_code is not None:
out = [r for r in out if r.get("industry_code") == industry_code]
if agglvl_code is not None:
out = [r for r in out if r.get("agglvl_code") == agglvl_code]
return out

View File

@ -2,7 +2,7 @@
Pre-built wage and compensation series IDs — OES, ECI, ECEC, CPS earnings.
"""
from ..series import oes_national, eci
from ..series import oes_national, eci, ecec
# ---------------------------------------------------------------------------
# OES — Occupational Employment & Wage Statistics
@ -89,19 +89,58 @@ def eci_dashboard() -> dict:
# ---------------------------------------------------------------------------
# ECEC — Employer Costs for Employee Compensation
# ---------------------------------------------------------------------------
def ecec_total_compensation() -> str:
"""ECEC — civilian workers, total compensation cost per hour worked."""
return "CMU1010000000000D"
def ecec_total_compensation(owner: str = "civilian") -> str:
"""ECEC — total compensation cost per hour worked."""
return ecec("total_compensation", owner)
def ecec_total_benefits() -> str:
"""ECEC — civilian workers, total benefits cost per hour worked."""
return "CMU1036000000000D"
def ecec_wages_and_salaries(owner: str = "civilian") -> str:
"""ECEC — wages and salaries cost per hour worked."""
return ecec("wages_and_salaries", owner)
# Note: ECEC benefit subcomponents (health insurance, retirement & savings, etc.)
# are encoded as specific benefit-subcell codes in the series ID (suffix stays "D").
# Look them up against the ECEC component list before adding helpers — do not
# fabricate them with a letter suffix (the old "...H"/"...R" forms were invalid).
def ecec_total_benefits(owner: str = "civilian") -> str:
"""ECEC — total benefits cost per hour worked."""
return ecec("total_benefits", owner)
def ecec_paid_leave(owner: str = "civilian") -> str:
"""ECEC — paid leave cost per hour worked (vacation, holiday, sick, personal)."""
return ecec("paid_leave", owner)
def ecec_supplemental_pay(owner: str = "civilian") -> str:
"""ECEC — supplemental pay cost per hour worked (overtime, bonuses, shift diff)."""
return ecec("supplemental_pay", owner)
def ecec_health_insurance(owner: str = "civilian") -> str:
"""ECEC — health insurance cost per hour worked."""
return ecec("health_insurance", owner)
def ecec_retirement(owner: str = "civilian") -> str:
"""ECEC — retirement & savings cost per hour worked (defined benefit + contribution)."""
return ecec("retirement_and_savings", owner)
def ecec_legally_required(owner: str = "civilian") -> str:
"""ECEC — legally required benefits per hour (Social Security, Medicare, UI, workers' comp)."""
return ecec("legally_required", owner)
def ecec_dashboard(owner: str = "civilian") -> dict:
"""The compensation cost breakdown — what an hour of labor costs an employer."""
return {
"Total Compensation": ecec_total_compensation(owner),
"Wages & Salaries": ecec_wages_and_salaries(owner),
"Total Benefits": ecec_total_benefits(owner),
"Paid Leave": ecec_paid_leave(owner),
"Supplemental Pay": ecec_supplemental_pay(owner),
"Health Insurance": ecec_health_insurance(owner),
"Retirement & Savings": ecec_retirement(owner),
"Legally Required": ecec_legally_required(owner),
}
# ---------------------------------------------------------------------------

View File

@ -336,6 +336,55 @@ def eci(
return f"CI{adj}{owner}{component}{'0'*9}{estimate}"
# ---------------------------------------------------------------------------
# ECEC — Employer Costs for Employee Compensation
# Component ("estimate") codes from the BLS cm.estimate decode table.
# ---------------------------------------------------------------------------
_ECEC_COMPONENT = {
"total_compensation": "01",
"wages_and_salaries": "02",
"total_benefits": "03",
"paid_leave": "04",
"supplemental_pay": "09",
"insurance": "13",
"health_insurance": "15",
"retirement_and_savings": "18",
"legally_required": "21",
"workers_compensation": "27",
}
_ECEC_OWNER = {
"civilian": "1",
"private": "2",
"state_local_gov": "3",
}
def ecec(
component: str = "total_compensation",
owner: str = "civilian",
datatype: str = "D", # D = cost per hour worked (dollars)
) -> str:
"""
ECEC series — employer cost per hour worked, all industries / all workers.
Format: CM + U + owner(1) + component(2) + 10 zeros + datatype(1) = 17 chars.
Args:
component: key in _ECEC_COMPONENT (e.g. "health_insurance") or a raw 2-digit code
owner: "civilian" | "private" | "state_local_gov" (or raw "1"/"2"/"3")
datatype: "D" = cost per hour in dollars (default); "P" = percent of total comp
Examples:
ecec() → "CMU1010000000000D" (civilian total compensation)
ecec("health_insurance") → "CMU1150000000000D"
ecec("retirement_and_savings") → "CMU1180000000000D"
ecec("total_benefits", "private") → "CMU2030000000000D"
"""
own = _ECEC_OWNER.get(owner, owner)
code = _ECEC_COMPONENT.get(component, component)
return f"CMU{own}{code}{'0'*10}{datatype}"
# ---------------------------------------------------------------------------
# Productivity (Major Sector)
# ---------------------------------------------------------------------------