Python computation formulas for Treasury data analysis. Use for percent change, CAGR, stdev, kurtosis, VaR, regression, Theil index, and other statistical calculations.

2026-04-083

cpi-reference

jwalin-shah/officeqa-arena

CPI-U annual averages (1929-2024) for inflation adjustment. Use when question mentions inflation, CPI, constant dollars, or real dollars.

2026-04-083

data-files

jwalin-shah/officeqa-arena

How to read Treasury Bulletin data files in /app/resources/. Use when exploring resources or when JSON parsing fails.

2026-04-083

fiscal-calendar

jwalin-shah/officeqa-arena

U.S. fiscal year boundaries and calendar year conversions. Use when question mentions fiscal year, FY, or when you need to determine which months belong to a fiscal year.

2026-04-083

table-extraction

jwalin-shah/officeqa-arena

How to accurately extract values from Treasury Bulletin tables. Use when reading tables with many columns, similar row labels, or when you need to verify you have the right cell.

2026-04-083

import re,sys,difflib def clean(v): v=re.sub(r'\s*[0-9]+/\s*$','',v) # strip footnotes like 3/ v=re.sub(r'^[rp]/\s*','',v) # strip r/ p/ prefixes m=re.match(r'^$([0-9,.]+)$$',v) # (123) = -123 if m:v='-'+m.group(1) return v.strip() def to_num(v): v=clean(v).replace(',','').replace('$','').replace('%','') try:return float(v) except:return None def parse(lines): p=sum(l.count('|') for l in lines);t=sum(l.count('\t') for l in lines) d='|' if p>t else '\t' rows=[] for n,l in enumerate(lines): pts=[c.strip() for c in l.split(d) if c.strip()] if len(pts)>=2:rows.append((n,pts)) if not rows:return [],[],'' h=rows[0][1];data=rows[1:] units='' for l in lines[:15]: m=re.search(r'(millions?|billions?|thousands?|percent)',l,re.I) if m:units=m.group(0).lower();break return h,data,units args=sys.argv[1:];fp=args[0] rf=cf=lr=None;sr=sc=fu=False;i=1 while i<len(args): if args[i]=='--row':rf=args[i+1].lower();i+=2 elif args[i]=='--col':cf=args[i+1].lower();i+=2 elif args[i]=='--lines':m=re.match(r'(\d+)-(\d+)',args[i+1]);lr=(int(m[1])-1,int(m[2])) if m else None;i+=2 elif args[i]=='--rows':sr=True;i+=1 elif args[i]=='--cols':sc=True;i+=1 elif args[i]=='--fuzzy':fu=True;i+=1 else:i+=1 with open(fp,errors='replace') as f:al=f.readlines() lines=al[lr[0]:lr[1]] if lr else al h,data,units=parse(lines) if not h:print(''.join(lines[:30]));sys.exit() if units:print(f'[units: {units}]') if sc: for j,c in enumerate(h):print(f'[{j}] {c}') elif sr: for n,pts in data:print(f'[{n+1}] {pts[0]}') else: for n,pts in data: label=pts[0] if rf: if fu: ratio=difflib.SequenceMatcher(None,rf,label.lower()).ratio() if ratio<0.5:continue elif rf not in label.lower():continue for j in range(1,len(pts)): cn=h[j] if j<len(h) else f'c{j}' if cf and cf not in cn.lower():continue raw=pts[j] if j<len(pts) else '' if not raw or raw=='nan':continue num=to_num(raw) if num is not None:print(f'{label} | {cn} = {num}') else:print(f'{label} | {cn} = {clean(raw)}')

import re,sys,os q=sys.argv[1];d=sys.argv[2] if len(sys.argv)>2 else '/app/resources' for fp in sorted(os.listdir(d)): if not fp.endswith('.txt'):continue with open(os.path.join(d,fp),errors='replace') as f:lines=f.readlines() for i,l in enumerate(lines): if re.search(q,l,re.I): ctx=''.join(lines[max(0,i-1):i+2]).rstrip() print(f'[{fp}:{i+1}]\n{ctx}\n')

python3 /tmp/e.py FILE --cols # list columns python3 /tmp/e.py FILE --rows # list rows with line numbers python3 /tmp/e.py FILE --row "defense" --col "1953" # exact cell python3 /tmp/e.py FILE --row "defense" # all cols for row python3 /tmp/e.py FILE --row "natl def" --fuzzy # fuzzy match python3 /tmp/g.py "keyword" /app/resources # search files python3 /tmp/batch.py "metric" /app/resources # extract row from ALL files at once

python3 /tmp/e.py FILE --rows # scan all row labels — find the section you need grep -n "Table\|TABLE" FILE # find table boundaries by line number python3 /tmp/e.py FILE --lines START-END --row "target" # extract within that section only

import re,sys,os row=sys.argv[1];d=sys.argv[2] if len(sys.argv)>2 else '/app/resources' for fp in sorted(os.listdir(d)): if not fp.endswith('.txt'):continue full=os.path.join(d,fp) with open(full,errors='replace') as f:lines=f.readlines() for l in lines: if re.search(row,l,re.I): vals=re.findall(r'[\d,]+\.?\d*',l) if vals:print(f'{fp}: {" | ".join(vals)}') break

officeqa-tools

Save to /tmp/e.py — robust table extractor:

Save to /tmp/g.py — grep with context:

Usage:

For tables with multiple sections:

Multi-file batch extraction (save as /tmp/batch.py):

Precision report — use for any "rounded to N places" question:

Computation recipes — what data you need:

Save to /tmp/e.py — robust table extractor:

Save to /tmp/g.py — grep with context:

Usage:

For tables with multiple sections:

Multi-file batch extraction (save as /tmp/batch.py):

Precision report — use for any "rounded to N places" question:

Computation recipes — what data you need:

name	officeqa-tools
description	Table parser + computation recipes — save scripts to /tmp