# NSE daily market data pack

Indian equities from the National Stock Exchange, cleaned and compiled into parquet files that
open in seconds. Built 2026-09-23.

## What is in it

| File | Rows | Symbols | Covers |
|---|---|---|---|
| `daily.parquet` | 10,817,987 | 5,482 | 1994-11-03 to 2026-09-22 |
| `indices.parquet` | 293,348 | 176 indices | 2012-02-21 to 2026-09-22 |
| `market_cap.parquet` | 3,637,218 | 2,422 | 2015-01-28 to 2026-08-25 |
| `corporate_actions.csv` | 1,355 | splits, bonuses, rights, demergers | |
| `dividends.csv` | 23,994 | dividends with ex-dates | |
| `sample_RELIANCE_daily.csv` | 7,927 | one stock, plain CSV | |

## Columns

`daily.parquet`: symbol, series, date, open, high, low, close, prevclose, volume, turnover

`indices.parquet`: symbol, name, date, open, high, low, close, volume, turnover, pe, pb, divyield

`market_cap.parquet`: symbol, date, close, shares_outstanding, market_cap, shares_source

Prices are in rupees, as the exchange published them on the day. Volume is shares, turnover is
rupees. Dates are calendar dates, one row per symbol per trading session.

## Adjusting for splits and bonuses

Prices are unadjusted, which is what the exchange publishes. `corporate_actions.csv` carries the
ex-date and the ratio for each split, bonus, rights issue and demerger, so you can build whatever
adjustment your method wants rather than inheriting someone else's. Multiply every price before
an ex-date by the factor in that row to get a continuous series. `dividends.csv` is separate
because dividend adjustment is a choice, not a fact.

## Reading it

```python
import pandas as pd
daily = pd.read_parquet("daily.parquet")
reliance = daily[daily.symbol == "RELIANCE"].set_index("date").sort_index()
```

Parquet reads in pandas, polars, DuckDB, R (arrow) and anything else that speaks Arrow. DuckDB
will query the file without loading it:

```sql
SELECT symbol, max(close) FROM 'daily.parquet' WHERE date > '2020-01-01' GROUP BY symbol;
```

## Where it comes from, and what you may do with it

Source: the National Stock Exchange of India's own published files, the daily bhav copies, the
index close files and the corporate action feeds. NSE is the source of the underlying data and
asserts rights over it. This pack is the cleaning, gap filling and compilation work, not a
licence to the data. Check NSE's data licensing terms before redistributing it, embedding it in
a product you sell, or using it commercially.

No warranty. Exchange data carries errors, and so does any cleaning of it. Check anything you
intend to trade on.

## Known limits

- Unadjusted prices, by design. See above.
- Market caps start 2015-01-28, because shares outstanding are not in the bhav copies.
- A delisted stock's series stops on its last trading day and is kept.
- Symbol changes are not stitched: a company that changed its NSE symbol appears under both.
