THULab/indian_markets

TejHQ Indian Markets (TsFile)

Apache TsFile version of tejhq/indian-markets.

Overview

End-of-day data for every NSE and BSE listed equity, built straight from the exchanges' official bhavcopy. The source repository is a bundle of related tables: raw prices, corporate actions, back-adjusted prices, symbol history, derived metrics, and a survivorship-bias-free liquidity universe. It is refreshed every trading day at 20:00 IST by an open pipeline; the same data is served at api.tejhq.dev.

Each logical family is converted into its own TsFile table:

TsFile table	Rows	Source tree
prices_raw	8,448,626	nse/year=YYYY/, bse/year=YYYY/ (raw bhavcopy)
prices_adjusted	8,448,626	prices_adjusted/ (back-adjusted OHLCV)
metrics	8,448,640	metrics/ (returns + rolling stats)
actions	41,836	actions/ (corporate actions)
universe	108,500	universe/ (point-in-time liquidity membership)
symbol_history	7,995	symbol_history/ (symbol/ISIN validity windows)

• Coverage: NSE 2010-01-04 to present (~4,100 trading days, symbols in series EQ/BE/BZ); BSE 2024-07-08 to present (series A/B/T). 7,339 (exchange, symbol) instrument devices across the price tables.
• Frequency: daily (one end-of-day record per instrument per trading day).
• Record meaning: one instrument's end-of-day market data / metric / corporate-action / index-membership row, keyed by (exchange, symbol, Time).

Schema (TsFile structure)

Every table uses milliseconds since the Unix epoch as Time and stores exchange and symbol as TAG columns (a device is one (exchange, symbol) instrument):

• prices_raw — Time = trading date; TAGs exchange, symbol; FIELDs series, isin, name, open, high, low, close, last, prev_close, volume, turnover, trades, year, month. Raw bhavcopy rollups from the nse/ and bse/ trees.
• prices_adjusted — Time = trading date; TAGs exchange, symbol; FIELDs as above plus adj_factor_cumulative and adj_close (continuous through splits, bonuses and dividends).
• metrics — Time = trading date; TAGs exchange, symbol; FIELDs isin, adj_close, ret_1d, ret_5d, ret_21d, ret_63d, ret_126d, ret_252d, ret_ytd, high_52w, low_52w, pct_off_52w_high, pct_off_52w_low, avg_vol_20d, avg_vol_60d, avg_turnover_20d. Rolling fields are null until a full window of history exists.
• actions — Time = ex_date; TAGs exchange, symbol; FIELDs isin, company, record_date, type, ratio_num, ratio_den, cash_amount, face_value_from, face_value_to, raw_subject.
• universe — Time = rebalance_date (first trading day of each month); TAGs exchange, symbol; FIELDs valid_to, rank, isin, name, avg_turnover_63d. This lookup table has no separate event timestamp, so the monthly rebalance date is used as Time.
• symbol_history — Time = valid_from (start of each symbol/ISIN validity window); TAGs exchange, symbol; FIELDs isin, valid_to, trading_days. This lookup table has no separate event timestamp, so the window start valid_from is used as Time.

String-valued source columns are stored as STRING FIELDs; null strings become empty strings. trades (null before 2012 on NSE) and the nullable action ratio/face-value columns are stored as DOUBLE because the source files represent them as nullable floats.

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("actions.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())

Source & license

• Original dataset: https://huggingface.co/datasets/tejhq/indian-markets
• Author / publisher: TejHQ (https://tejhq.dev)
• Paper: none; pipeline source code: https://github.com/tejhq/tej-bazaar (MIT)
• License: MIT (pipeline); the underlying data is exchange-published NSE/BSE EOD bhavcopy and corporate-action feeds, redistributed as cleaned parquet. Verify exchange terms before commercial use.