Dataframe schemas
These are the unselected Arrow/Polars schemas from
src/marketdata/storage/schema.py. Expressions supplied to GetDataframe may
change the returned columns. Runtime quote output is separate from the full
instrument catalogue.
Base frame
Section titled “Base frame”| Column | Polars type | Meaning |
|---|---|---|
isin |
String | Instrument identifier |
lastPrice |
Int64 | Last trade price |
lastPricePct |
Float32 | Last-price percentage change supplied by the source |
closePrice |
Int64 | Closing price |
closePricePct |
Float32 | Closing-price percentage change supplied by the source |
limits |
Array(Struct, configured depth) | Fixed-depth book rows |
Prices use source/native units. Missing data remains nullable. A missing quote or count is not evidence of zero liquidity.
Option frame
Section titled “Option frame”The option frame contains every base column plus:
| Column | Polars type | Meaning |
|---|---|---|
buyIv |
Float64 | Calculated bid-side implied volatility |
sellIv |
Float64 | Calculated ask-side implied volatility |
blackScholesPrice |
Float64 | Calculated theoretical option price |
Calculated values depend on instrument metadata, available history and current
quotes. Missing or invalid inputs can produce null model prices or NaN implied
volatilities; check for both null and non-finite values. The
configured risk_free_rate defaults to 0.3.
Nested book rows
Section titled “Nested book rows”limits has fixed depth, controlled by market_data.dataframe_max_limits_rows
(default five). Rows are stored in best-to-deeper order. Each struct contains:
| Field | Polars type | Meaning |
|---|---|---|
buyPrice, sellPrice |
Int64 | Bid/ask prices |
buyVolume, sellVolume |
Int64 | Bid/ask volumes |
buyCount, sellCount |
Int64 | Order counts when supplied by the source |
timestamp |
Datetime(us) | Book-row time; no timezone annotation in the Arrow schema |
buyValue, sellValue |
Float64 | Price multiplied by volume where available |
Book timestamps and snapshot capture timestamps are distinct. Repeated snapshots do not establish that each stored row was refreshed. Preserve the timezone semantics of the producing source when interpreting a timestamp without a timezone annotation; do not silently assign the consumer host’s timezone.
Recorded snapshots
Section titled “Recorded snapshots”Persistence schemas add the following fields to the base/option schema:
| Column | Polars type | Meaning |
|---|---|---|
snapshot_id |
String | Snapshot identifier |
captured_at |
Datetime(us, UTC) | Capture time |
market_type |
UInt8 | Market discriminator |
These metadata columns belong to recorded snapshots, rather than the default
GetDataframe Arrow frames. The gRPC response carries capture time separately in
GetDataframeResponse.timestamp.
See the Market Data protobuf reference for the
separate scalar wire types used by GetBestLimits.