Skip to content

Reads a column holding JSON text and expands it into output rows — one per value found in the document. The counterpart to Table XML Parse, and the PlaidCloud equivalent of the Alteryx JSON Parse tool.

Use it when an upstream system hands you a JSON payload in a single column and you need its contents as ordinary rows and columns.

nested JSON{“id”: 1,“user”: { “name”:“Ann” },“tags”: [“a”,“b”]}parseflat columnsiduser.nametags1Anna, b
Table JSON Parse flattens nested JSON into columns — nested objects become dotted paths (user.name) and arrays are expanded, so downstream steps work in plain rows and columns.

The column holding the JSON text to parse.

  • Name and value only (default) — each output row carries JSON_Name and JSON_ValueString. Simplest, and enough when everything downstream treats the values as text.
  • Typed values — turn the single-field option off to add the type-specific value columns alongside, so numbers and booleans arrive as numbers and booleans rather than strings.

Off by default. Turn it on to repeat the input record’s other columns on every value row, so parsed values stay joined to the record they came from without a separate join step.

Output (json_layout) picks a different output shape. Each of these keeps one row per input row:

  • Wide: one column per key path (wide) — the input columns, less the JSON field unless it is included, plus one text column per dotted key path, NULL where a document lacks it.
  • One column per JSONPath query (paths) — the input columns plus one column per JSONPath query. List the queries in JSONPath Queries (json_paths), each with a Column, JSONPath, Type and Several Matches setting (target, path, dtype, multi): path is a JSONPath (names, indexes, wildcards, deep scan, slices, unions, and filters), dtype is text, integer, bigint, double, boolean or json, and multi is empty for the first match (NULL when none), list for every match as a list value, or paths for the paths that match. Instead of the list, Queries from a Table (json_dictionary) reads the queries from the rows of a table: {source, path, type, output} name the table and its columns holding each path, type and output column name.
  • Columns as KNIME JSON to Table makes them (table) — the input columns plus the columns KNIME’s JSON to Table node makes of the documents. The JSON to Table options (json_table) set it, starting with Name columns by their JSON path (paths), Path Separator (separator, default .) to join the path segments, Arrays (arrays) as GenerateCollectionCells, GenerateColumns or KeepAllArrayAsJsonArray, Levels to Expand (levels, 0 for all), and Omit objects that were expanded (omit_objects, default true) to leave out object values.

With the JSONPath or JSON to Table layout, turn on Leave the JSON field out of the output (json_remove_source) to drop the JSON field from the result.