twilight venturesopen a parquet file in the IDE
A columnar file keeps its values by column and its map in a footer at the end, so the schema can be read before a single row is touched. Parquet Lens opens a .parquet file in an editor tab, straight off disk, with no notebook and no conversion step. It runs in IntelliJ IDEA, PyCharm and every other JetBrains IDE on 2024.2 or newer, Android Studio from Ladybug.
These files come out of training data, eval sets, feature tables and lakehouse exports. The same plugin opens .avro containers, TFRecord datasets plain or compressed, and Arrow IPC and Feather v2 files. Nothing is loaded whole, and a 10 GB dataset pages the way a 10 MB one does.
installing and opening
Install Parquet Lens from the JetBrains Marketplace and open the file from the project tree. The Schema tab is up before a row decodes, carrying the logical and physical type of every column, its codec and encodings, and the null count with the minimum and maximum the footer stores per column. Rows then arrive in pages of up to 5000, go-to-row lands on any row you name, and Ctrl+F streams a text search across every row.
the values you will see
Timestamps keep their unit and timezone, decimals and int64 keep their precision, and a nested list or struct or map renders as JSON with a pretty detail pane beside it. We wrote the Arrow IPC reader ourselves against the columnar format spec, and the test suite validates every decode path against files written by the reference implementation. A corrupt or truncated file reports a clear reason, and a codec the reader does not have is refused by name.
free, and what Pro adds
Reading is free. A row copies out as JSON and the page as CSV. Pro is a subscription with a 30-day trial. It adds a filter such as lang == "en" && score >= 0.5 evaluated across every row group with min/max pruning where the writer stored those statistics, a sort of the whole dataset by any column, exact distinct counts over every row, and an export of all rows or the matches to CSV, JSONL or a new Parquet file. The product page lists both tiers and quotes the price as the store shows it.
The same files open in Parquet Lens for VS Code and, as an attachment, on a Confluence page or a Jira issue with Lens File Viewer, and the guide on columnar files puts all three side by side.