Table Cast Type¶
This node converts a single column in a table to a different data type: string, int, float, bool, or datetime. It lets you control how unparseable values are treated, either coercing them to null, raising an error, or dropping the entire row. The node returns the updated table, the number of failed conversions, and both human-readable and structured logs for downstream inspection or monitoring.
Usage¶
Use this node when you need to standardize or correct column data types before analytics, joins, aggregations, or charting. Common scenarios include converting numeric-looking strings from CSV imports into real numbers, turning yes/no or 0/1 fields into booleans, and parsing timestamp strings into datetime objects for time-based grouping or filtering. It typically appears downstream of ingestion nodes such as SaltTableFromCSV or SaltTableFromJSON and upstream of transform/analysis nodes like SaltTableFilter, SaltTableGroupBy, SaltChart, SaltTableInfo, and SaltTableOutput.
Choose target_type based on how the column will be used: use float or int for metrics, bool for flags, datetime for time-series, and string for free text. Use error_strategy="coerce_null" to keep all rows while marking invalid values as null, "raise" when you need strict data quality and want the workflow to fail fast, and "drop_row" when rows with invalid values should be excluded from further processing. When converting to datetime, provide datetime_format if the input format is known and consistent (for example %Y-%m-%d %H:%M:%S) to improve accuracy and performance. Combine this node with SaltTableFillNull to repair missing or coerced-null values, and with SaltTableInfo to verify resulting data types and null counts.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| table | True | TABLE | Input tabular data to modify. Can be a Salt table or a DataFrame-like structure. Must contain the column specified in column. Only the selected column is cast; all other columns are preserved as-is. | A table with columns: ["user_id" (string), "signup_date" (string), "is_active" (string), "monthly_spend" (string)]. |
| column | True | STRING | Name of the column to cast to a new type. Must exactly match an existing column name. If the column is missing, execution fails with an error listing the available columns. | monthly_spend |
| target_type | True | STRING enum | Target data type for the specified column. Allowed values: "string", "int", "float", "bool", "datetime". string converts all values to text; int and float convert to numeric values; bool maps common true/false representations; datetime parses date or datetime strings. | float |
| error_strategy | True | STRING enum | Controls how values that cannot be parsed into the target type are handled. Allowed values: "coerce_null" (unparseable values become null, all rows kept), "raise" (the node raises an error as soon as an invalid value is found), "drop_row" (rows containing unparseable values in the target column are removed from the output table). | coerce_null |
| datetime_format | False | STRING | Optional strptime-style format string used only when target_type is "datetime". When provided, it guides datetime parsing (for example %Y-%m-%d or %Y-%m-%d %H:%M:%S). When omitted, automatic parsing is used, which may be slower or ambiguous for certain formats. | %Y-%m-%d %H:%M:%S |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| table | TABLE | The resulting table after casting the specified column to the selected type. Other columns are unchanged. Depending on error_strategy, some rows may be dropped (drop_row) or contain nulls where values could not be parsed (coerce_null). | A table where "monthly_spend" is now a numeric float column [19.99, 0.0, null, 42.5] instead of strings like ["19.99", "0", "N/A", "42.5"]. |
| error_count | INT | Number of values in the target column that failed to parse into the requested type. Counts only cases where a previously non-null value became null because of conversion failure. | 42 |
| log | STRING | Human-readable summary of the cast operation, including the column name, target type, number of failed parses, and the selected error strategy. | Cast [monthly_spend] to float: 42 values failed (coerce_null) |
| log_json | STRING | Structured JSON string describing the operation, including node name, input and output row counts, parameters (column, target_type, error_strategy, datetime_format if set), and error_count. Designed for logging, monitoring, and audit pipelines. | {"node":"SaltTableCastType","input_rows":10000,"output_rows":9975,"parameters":{"column":"monthly_spend","target_type":"float","error_strategy":"drop_row"},"error_count":25} |
Important Notes¶
- Performance: Casting large columns, especially to datetime, can be slow and memory-intensive; providing a precise datetime_format often improves performance and reliability.
- Limitations: Bool casting only recognizes common textual representations such as true/false, 1/0, yes/no, t/f, y/n (case-insensitive); other non-null values become null or trigger row drops or errors depending on error_strategy.
- Behavior: With error_strategy="drop_row", the output table may have fewer rows than the input, which can impact downstream joins, time-series alignment, or aggregations expecting stable row counts.
- Behavior: When casting to int, the column uses a nullable integer type so existing nulls and failed parses remain as null instead of being forced to 0, preserving information about missing or invalid data.
- Behavior: When casting to string, the node preserves actual missing values as null rather than turning them into literal strings like "nan" or "None".
Troubleshooting¶
- Error: Column 'X' not found. Available: ...: The specified column does not exist in the input table. Check spelling and capitalization, or inspect the upstream table schema using a node like SaltTableInfo to confirm column names.
- Error: Cannot convert 'value' to float/int/bool/datetime in column 'X'.: This occurs when error_strategy is "raise" and at least one value is invalid for the chosen target_type. Either clean the data upstream (for example with SaltTableFilter or SaltTableFillNull) or switch error_strategy to "coerce_null" or "drop_row".
- Many more nulls after casting than expected: If error_strategy is "coerce_null" and many values become null, verify that target_type is appropriate and that datetime_format (if used) matches the actual data format. Inspect a sample of the raw column values to adjust your settings or preprocess the data.
- Unexpected reduction in row count: If the output has fewer rows, error_strategy is likely set to "drop_row" and some values could not be parsed. Use error_count and log_json to see how many rows were affected, or switch to "coerce_null" if preserving all rows is more important than dropping bad ones.