PakDataKit for Python
Utilities for Pakistani data: CNIC, mobile numbers and networks, WhatsApp links, IBAN, postal codes, cities and districts, address parsing, Roman Urdu, Hijri dates, public holidays, rupee formatting, amounts in words, NTN/STRN, landlines and Raast QR codes.
This is the Python port of the pakdatakit npm package.
It ships the same datasets and is tested against shared test vectors
generated from the TypeScript library, so both give identical results.
Pure Python, no dependencies, Python 3.9+.
Install
pip install pakdatakit
Examples
import datetime as dt
import pakdatakit as pdk
pdk.validate_cnic("35202-1234567-1") # True
pdk.get_cnic_info("35202-1234567-1")
# CNICInfo(digits='3520212345671', formatted='35202-1234567-1', province='Punjab', gender='male')
pdk.format_phone("0300 1234567") # '+923001234567'
pdk.get_network("0345-1234567") # 'Telenor'
pdk.create_whatsapp_link("03001234567", "Salam!")
pdk.validate_iban("PK36 SCBL 0000 0011 2345 6702") # True
pdk.get_bank_from_iban("PK36SCBL0000001123456702") # 'Standard Chartered Bank (Pakistan) Limited'
pdk.get_bank_info("MEZN").type # 'islamic'
pdk.get_province("Lahore") # 'Punjab'
pdk.get_divisions("Sindh") # ['Hyderabad', 'Karachi', 'Larkana', ...]
pdk.get_division_of_district("Kasur") # 'Lahore'
pdk.get_tehsils("Rawalpindi") # ['Gujar Khan', 'Kahuta', ..., 'Murree', ...]
pdk.get_tehsil_district("Murree") # 'Rawalpindi'
pdk.parse_address("House 12, Street 5, Model Town Lahore 54000")
pdk.get_city_by_postal_code("54660") # 'Lahore'
pdk.get_post_office("48330")
# PostOffice(code='48330', office='Dulmial', town='Dulmial', district=None, province='Punjab')
pdk.format_pkr(150000) # 'Rs. 150,000'
pdk.format_lakh(2500000) # '25 Lakh'
pdk.amount_in_words(125000, cheque=True) # 'Rupees One Lakh Twenty-Five Thousand Only'
pdk.number_to_words(125000, language="ur") # 'ایک لاکھ پچیس ہزار'
pdk.format_hijri(dt.date(2026, 3, 1)) # '12 Ramadan 1447'
pdk.get_pakistan_holidays(2026)[0] # Holiday(id='kashmir-day', ...)
payload = pdk.create_raast_p2p_qr("PK36SCBL0000001123456702", amount=1500, expires_in="1d")
pdk.validate_raast_qr(payload).valid # True
pdk.format_cnic_input("3520212") # '35202-12'
pdk.check_phone_input("0300 12") # 'incomplete'
API
Every function of the npm package is available under its snake_case name
(validateCNIC → validate_cnic, createRaastP2PQr → create_raast_p2p_qr).
Option objects become keyword arguments ({ expiresIn: '1d' } →
expires_in="1d"), structured results are frozen dataclasses with snake_case
fields, null is None, and a RangeError is a ValueError. Dates are
datetime.date / naive local datetime.datetime.
Geography:
| Function | Returns |
|---|---|
get_province(city) |
Province of a city, or None |
search_cities(query) |
Matching city names, sorted |
get_district_province(district) |
Province of a district, or None |
search_districts(query) |
Matching district names, sorted |
get_districts_by_province(province) |
Districts of a province, sorted |
get_divisions(province=None) |
The 30 divisions of the 2023 census, or those of one province, sorted |
get_division_of_district(district) |
Division of a 2023 census district, or None |
get_tehsils(district) |
Tehsils (talukas, sub-divisions) of a district, sorted |
get_tehsil_district(tehsil, province=None) |
Census district of a tehsil, or None if unknown or ambiguous |
Divisions and tehsils come from the 2023 census. Each of the 591 tehsils is
recorded under the district it was in at the census, so get_tehsils and
get_tehsil_district use census districts (Rawalpindi still lists Murree,
and a tehsil of Quetta is reported under Quetta). Districts created after
the census have no division and may have no tehsils. Azad Jammu and Kashmir
and Gilgit-Baltistan are not covered.
Postal codes:
| Function | Returns |
|---|---|
get_city_by_postal_code(code) |
City of a post code, or None (only cities on pakdatakit's city list) |
get_postal_code(city) |
Primary (GPO) post code of a city, or None |
get_postal_codes(city) |
All post codes of a city, sorted |
get_postal_areas(city) |
Post office names of a city, sorted |
search_postal_codes(query) |
City post codes matching a city, office name or code prefix, sorted |
get_post_office(code) |
PostOffice for any delivery post office, or None |
search_post_offices(query) |
PostOffices whose office name or town contains the query, or whose code starts with it, sorted by code |
get_city_by_postal_code only answers for post offices in a city on
pakdatakit's city list (54660 is Lahore); the code of an office in another
town or village returns None. get_post_office and search_post_offices
cover every delivery office in Pakistan Post's directory (about 2,560),
including towns and villages, as PostOffice(code, office, town, district, province). district is set only when the town is the headquarters a
district is named after.
Delivery checks:
| Function | Returns |
|---|---|
score_address(address) |
AddressScore(score, missing, parsed): how complete a delivery address is |
phone_risk_signals(phone) |
Reasons to double-check a delivery phone number; [] for an ordinary mobile |
score_address builds on parse_address and adds up ADDRESS_WEIGHTS for
the parts it finds: city 30, houseNumber 20, street 15, area 10,
postalCode 15 (5 for a code that is not a known post office) and landmark
10 ("near", "opp", "nazd", "qareeb", "سامنے" ...). A post office in a different
province from the city takes 15 off and adds a warning to parsed.warnings.
missing lists the parts not found, in that order. phone_risk_signals
returns any of invalid, landline, not-mobile, repeated-digits,
sequential-digits and test-number. Both are heuristics and never raise.
pdk.score_address("House 12, Street 4, Block B, Gulberg III, near Liberty Market, Lahore 54660").score # 100
pdk.score_address("Lahore").missing # ['houseNumber', 'street', 'area', 'postalCode', 'landmark']
pdk.phone_risk_signals("03001234567") # ['sequential-digits', 'test-number']
Sales tax:
| Function | Returns |
|---|---|
get_sales_tax_rate(kind=, province=None, date=None) |
SalesTaxRate for goods or a province's services on a date, or None |
calculate_sales_tax(amount, rate=, province=, kind=, date=, inclusive=False) |
SalesTaxAmounts(net, tax, gross, rate), or None |
split_inclusive(gross, rate) |
SalesTaxAmounts for an amount that already includes the tax, or None |
Standard rates only, from 1 July 2019: federal sales tax on goods (FBR) and
sales tax on services in Punjab, Sindh, Khyber Pakhtunkhwa, Balochistan and
Islamabad Capital Territory, each with its law and source.
SalesTaxRate has the npm package's fields, with from named from_.
province accepts KP, KPK, ICT and Islamabad; date is
YYYY-MM-DD or a datetime.date and defaults to today. calculate_sales_tax
uses rate if given, otherwise looks one up (kind defaults to services
when a province is given). Amounts are rounded half up to the paisa as
written (1.005 is Rs 1.01), and net + tax always equals gross. Amounts
from 0 to one trillion rupees and rates from 0 to 100 with at most two
decimals are accepted; anything else returns None, never an exception.
pdk.get_sales_tax_rate(kind="services", province="Sindh", date="2024-06-30").rate # 13
pdk.calculate_sales_tax(1000, rate=18) # SalesTaxAmounts(net=1000.0, tax=180.0, gross=1180.0, rate=18.0)
pdk.calculate_sales_tax(1150, province="Sindh", inclusive=True).net # 1000.0
pdk.split_inclusive(99.99, 18) # SalesTaxAmounts(net=84.74, tax=15.25, gross=99.99, rate=18.0)
Income tax on salary:
| Function | Returns |
|---|---|
calculate_salary_tax(annual_income, tax_year=None, date=None) |
SalaryTax(tax, monthly, effective_rate, slab, tax_year), or None |
calculate_monthly_salary_tax(monthly_salary, tax_year=, date=) |
The same for monthly_salary × 12; monthly is the tax to withhold a month |
get_salary_tax_slabs(tax_year=None) |
SalaryTaxYear(tax_year, from_, to, law, source, slabs), or None |
get_salary_tax_years() |
The tax years covered, oldest first ([2020, …, 2027]) |
get_tax_year(date=None) |
The tax year (July to June) a date falls in, or None for an invalid date |
The salaried rate table of the Income Tax Ordinance 2001 (First Schedule,
Part I, Division I, clause (2), which applies when salary is more than 75% of
taxable income), for tax years 2020 to 2027. A tax year runs from 1 July to
30 June and is named after the June it ends in. tax_year (an integer) takes
precedence over date (YYYY-MM-DD or a datetime.date); with neither, the
current tax year is used, and a year that is not covered returns None. Each
slab is SalaryTaxSlab(over, up_to, fixed, rate). Tax is rounded half up to
the paisa. Only the rate table is applied: the surcharge, tax credits,
exemptions, the separate rate for pensions and withholding adjustments are
not. Amounts from 0 to one trillion rupees are accepted; anything else
returns None, never an exception.
pdk.calculate_salary_tax(1_800_000, tax_year=2026)
# SalaryTax(tax=72000.0, monthly=6000.0, effective_rate=4.0,
# slab=SalaryTaxSlab(over=1200000, up_to=2200000, fixed=6000, rate=11), tax_year=2026)
pdk.calculate_monthly_salary_tax(150_000, tax_year=2026).monthly # 6000.0
pdk.get_salary_tax_slabs(2027).law # 'Income Tax Ordinance 2001, First Schedule, ...'
pdk.get_tax_year("2025-07-01") # 2026
Urdu text:
| Function | Returns |
|---|---|
slugify(text, script="any") |
A URL slug: lower-case words joined by -, Urdu letters kept as they are |
compare_urdu(a, b) |
-1, 0 or 1 in Urdu alphabetical order (CLDR 49), for cmp_to_key |
is_rtl(text) |
Whether the first letter is from a right-to-left script |
wrap_bidi(text, direction="auto") |
The text wrapped in Unicode directional isolates |
slugify drops diacritics and zero-width characters, writes the Arabic
look-alikes ك ي ى as ک ی ی, turns Urdu digits into 0-9 and folds accented
Latin-1 letters (é → e, ß → ss); script is any, latin or urdu
(SlugScript). compare_urdu follows the CLDR collation for Urdu (ا آ ب بھ پ
پھ … ی یھ ے, a letter joined with ھ is a letter of its own), the same in every
port without depending on ICU: digits come before letters one digit at a time
(10 before 2), and other scripts after. wrap_bidi uses FSI … PDI, or
RLI/LRI with "rtl"/"ltr" (BidiDirection). None of these raise; a value
that is not a string gives "" (or False).
pdk.slugify("Lahore Fort, 2024") # 'lahore-fort-2024'
pdk.slugify("لاہور Fort", script="latin") # 'fort'
pdk.slugify("کراچی ۲۰۲۴", script="urdu") # 'کراچی-2024'
sorted(["ے", "بھائی", "آم", "باغ", "اب"], key=functools.cmp_to_key(pdk.compare_urdu))
# ['اب', 'آم', 'باغ', 'بھائی', 'ے']
pdk.is_rtl("123 سلام") # True
f"آرڈر {pdk.wrap_bidi('PDK-1042')} بھیج دیا گیا"
FBR digital invoicing:
| Function | Returns |
|---|---|
build_fbr_invoice(input) |
FbrInvoiceBuild(invoice, totals) with the sales tax worked out, or None |
validate_fbr_invoice(invoice, today=None) |
FbrInvoiceValidation(valid, errors), each FbrInvoiceError(code, field, message) |
expected_sales_tax(item) |
The sales tax FBR expects on a line, in rupees, or None |
The invoice JSON for FBR's Digital Invoicing API (postinvoicedata and
validateinvoicedata), following PRAL's "Technical Specification for DI API"
version 1.12 (FBR_DI_SPEC_VERSION) and PRAL's Error Message Guide. The
invoice is a plain dict with FBR's camelCase keys (described by the
FbrInvoice and FbrInvoiceItem TypedDicts), ready for json.dumps; there
is no HTTP client and no token handling. build_fbr_invoice works out each
line's salesTaxApplicable when you leave it out (N% rates on the value
excluding sales tax, or the retail price for Third Schedule goods; Rs.N
rates on the quantity; other rates get 0), defaults totalValues,
fixedNotifiedValueOrRetailPrice and salesTaxWithheldAtSource to 0 and
buyerNTNCNIC and invoiceRefNo to "", and adds up the columns in
totals (a dict). validate_fbr_invoice returns every problem it can find
without calling FBR, with FBR's error code (None where FBR gives none):
required fields, formats (date, NTN/CNIC, FBR invoice number, sandbox
scenario), amounts (not negative, at most 2 decimals, 4 for quantity), SRO
rules and the sales tax arithmetic (to within one paisa). Pass
today="YYYY-MM-DD" to also reject future dates. Whether a province, HS code,
unit, rate, sale type or SRO is in FBR's lists is not checked. None of these
raise.
built = pdk.build_fbr_invoice({
"invoiceType": "Sale Invoice", "invoiceDate": "2025-04-21",
"sellerNTNCNIC": "1234567", "sellerBusinessName": "Example Traders",
"sellerProvince": "Sindh", "sellerAddress": "Karachi",
"buyerNTNCNIC": "7654321", "buyerBusinessName": "Example Buyer",
"buyerProvince": "Punjab", "buyerAddress": "Lahore",
"buyerRegistrationType": "Registered",
"items": [{"hsCode": "0101.2100", "productDescription": "Horses", "rate": "18%",
"uoM": "Numbers, pieces, units", "quantity": 2,
"valueSalesExcludingST": 1234, "saleType": "Goods at standard rate (default)"}],
})
built.invoice["items"][0]["salesTaxApplicable"] # 222.12
built.totals["salesTaxApplicable"] # 222.12
pdk.validate_fbr_invoice(built.invoice) # FbrInvoiceValidation(valid=True, errors=[])
pdk.validate_fbr_invoice({**built.invoice, "invoiceDate": "21-04-2025"}).errors
# [FbrInvoiceError(code='0005', field='invoiceDate', message='Invoice date must be in YYYY-MM-DD format')]
Utility bills:
| Function | Returns |
|---|---|
parse_electricity_ref(ref) |
ElectricityRef(type, digits, formatted, urban_rural, disco), or None |
validate_electricity_ref(ref) |
Whether the input is an electricity bill reference, customer ID or KE account |
validate_gas_consumer_number(number, company) |
Whether the input is an SNGPL (11 digits) or SSGC (10 digits) consumer number |
type is reference (DISCO reference number: 14 digits, optionally followed
by U or R), customer-id (DISCO customer ID: 10 digits, 11 for LESCO) or
k-electric (13 digits starting 04000). Spaces, dashes and dots are ignored
and Urdu digits are accepted. disco is K-Electric for a K-Electric account
and None otherwise: no official source maps DISCO digits to a distribution
company. No utility publishes a check digit, so only the format is checked.
company is SNGPL or SSGC (any case); anything else returns False.
None of these raise.
pdk.parse_electricity_ref("02-12624-0045901 R")
# ElectricityRef(type='reference', digits='02126240045901', formatted='02 12624 0045901 R', urban_rural='R', disco=None)
pdk.parse_electricity_ref("0400012345678").disco # 'K-Electric'
pdk.validate_electricity_ref("12345") # False
pdk.validate_gas_consumer_number("12345678901", "SNGPL") # True
Couriers:
| Function | Returns |
|---|---|
map_courier_status(courier, raw) |
A courier's status as one CourierStatus, or None |
normalize_city_for_courier(input) |
The city-list name for how a customer or courier writes a city, or None |
Experimental: map_courier_status and the CourierStatus values may
change in a minor release while more couriers' status lists are added. See
CONTRIBUTING.md to add one.
CourierStatus is one of pending, booked, picked-up, in-transit,
out-for-delivery, attempted, under-review, delivered, returning,
returned or cancelled. courier is TCS, Leopards, PostEx, M&P,
Trax or BlueEx (any case; LCS, MnP and Blue Ex are accepted), and
raw is the courier's status name, code or a non-negative integer ID (5,
5.0 and "5" are the same). Only statuses a courier documents publicly are
mapped: every order status and tracking code in PostEx's merchant API guide
(v4.1.9), and Leopards' "Returned to Shipper". TCS, M&P, Trax and BlueEx
publish no status list, so they return None, as does any unknown status.
normalize_city_for_courier accepts the name in any case, short forms
(LHR, Khi, Pindi, Isb, FSD), the old name Lyallpur, a trailing
"city" and the main cities in Urdu script. No API is called, and neither
raises.
pdk.map_courier_status("PostEx", "Out For Delivery") # 'out-for-delivery'
pdk.map_courier_status("PostEx", 5) # 'delivered'
pdk.map_courier_status("PostEx", "Attempt Made: Customer not available") # 'attempted'
pdk.map_courier_status("TCS", "Delivered") # None
pdk.normalize_city_for_courier("LHR") # 'Lahore'
pdk.normalize_city_for_courier("Karachi City") # 'Karachi'
pdk.normalize_city_for_courier("Atlantis") # None
BankInfo.type is one of commercial, public-sector, islamic,
microfinance, digital, emi, foreign, specialized or central.
BankInfo.active is False for a bank that merged into another or has closed.
DATA_VERSION gives the date the datasets were last checked, and
__version__ the package version.
See the main README for the full API reference and DATA.md for where each dataset comes from.
Development
From python/:
uv run --with pytest pytest -q
The data files in src/pakdatakit/data/ and the vectors in ../test-vectors/
are generated from the TypeScript package; don't edit them by hand. From the
repository root:
npm run build && node scripts/generate-test-vectors.mjs
License
MIT
Metadata
Release files for pakdatakit 2.5.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| pakdatakit-2.5.0.tar.gz | 129.2 kB | With provenance | Details for pakdatakit-2.5.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| pakdatakit-2.5.0-py3-none-any.whl | Python 3 | none | any | Details for pakdatakit-2.5.0-py3-none-any.whl |
Total release size: 267.8 kB
Release files / pakdatakit-2.5.0.tar.gz
| Download URL | pakdatakit-2.5.0.tar.gz |
|---|---|
| Size | 129.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
11908701b5d8eb92199496672a7911532b555e4dd1a21a911497a6677e847641
|
|
BLAKE2b-256 checksum How to use checksums |
da90070b07ae2f2509bc182e3f89f8282fd487ef3bc9bbd31c71dbafe444ad31
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 18, 2026.
Transparency logRelease files / pakdatakit-2.5.0-py3-none-any.whl
| Download URL | pakdatakit-2.5.0-py3-none-any.whl |
|---|---|
| Size | 138.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
5c4d283a88c9e160fe728c1837fed7f233d2099921dd1f4c1dc357de576b16d2
|
|
BLAKE2b-256 checksum How to use checksums |
74f9a17833a70d630eed362ef62cc72dc67b2e1520b07acd878c38bd7ce2264d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 18, 2026.
Transparency logSecurity for pakdatakit 2.5.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for pakdatakit-2.5.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 18, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.