TickerFlow — market-data investigation

Which market-data rows can you trust?

Investigate a suspicious CSV, inspect each validation decision, and follow accepted observations into usable time bars.

Investigation case

Preparing a bounded fixture…

—Input rows · same case

—Accepted

—Quarantined · unique rows

—Selected query rows

1. Inspect validation decisions

Choose a rule to reveal its affected rows. A row can break several rules; it is quarantined once. Duplicate pairs are both quarantined.

Rows to show

Row	Symbol	Original timestamp	Decision	Inspect

2. Query accepted data

All accepted observations make a temporary Parquet round trip and are queried with DuckDB. The run is cleaned up after the request; shared demo storage is untouched.

SymbolBar intervalStart included (UTC ISO)End excluded (UTC ISO)

Row	UTC timestamp	Symbol	Open	High	Low	Close	Volume

3. Trace rows into bars

Widths and labels show volume in synthetic units. Close prices have a separate label. Select a bar to inspect the actual contributing rows.

4. Reproduce the investigation

Price units: synthetic currency units. Volume: synthetic units. Timestamp precision: UTC microseconds. Symbols: uppercase. Data: unadjusted OHLCV, not trade ticks or order-book snapshots. No repairs or silent drops.

Validation rules · Quality regressions · Bar arithmetic tests