Tabular: pivot collection to wide

Connect a collection of keyed tabular datasets to collection_column_join when each element should contribute columns to one table. For two-column (identifier, value) elements, the result has one key column and one value column per element. The tool aligns keys across elements and fills cells where an element has no row for a key. Make the key unique within each input if the result must have one row per identifier.

Build a wide table

The checked-in example under verification/workflows/tabular-pivot-collection-to-wide/ has alpha and beta as elements of one dataset collection. Their tabular datasets have no header:

alpha         beta
K1    2        K2    5
K3    4        K3    7

Connect the collection to input_tabular and use this collection_column_join state:

tool_id: toolshed.g2.bx.psu.edu/repos/iuc/collection_column_join/collection_column_join/0.0.3
tool_state:
  input_tabular: { __class__: ConnectedValue }
  identifier_column: "1"
  has_header: "0"
  old_col_in_header: true
  fill_char: "0"
  include_outputs: null

The expected tabular_output, shown with spacing for readability, is:

#KEY  alpha_2  beta_2
K1    2        0
K2    0        5
K3    4        7

Here 0 means an absent key contributes a zero count. If an absent key means unknown rather than zero, choose a missing-value marker such as . instead. The file is tab-separated, and the Galaxy test compares its full contents with the committed expected file. The generated header uses #KEY for headerless inputs. With old_col_in_header: true, each value header combines the collection element identifier and the original column number, such as alpha_2. The MAPseq-to-ampvis2 workflow uses this headerless, zero-filled configuration.

Match the input and output contract

The Column join wrapper accepts multiple tabular inputs, including a dataset collection, and emits one tabular_output. Set its options from the actual files:

The two-column input is the simplest case. The wrapper also carries all non-key columns from each element into the output. For headered key/score inputs, use has_header: "1" and inspect the resulting column names. With old_col_in_header: false and elements alpha and beta, the output header is key, alpha, beta. The MAGs generation workflow repeatedly joins headered per-bin metric tables with fill_char: "." and old_col_in_header: false. Its output headers use element identifiers without the original column names.

Check the result before using it

• Check keys, values, and missing cells. Confirm that a key present in both elements aligns on one row and that a key present in only one element receives the chosen fill in the other column. A row-count or output-exists check alone will miss swapped values or scientifically wrong fills.
• Check key uniqueness and consistency. Each input should use the same key meaning and formatting. Duplicate keys within an element do not represent a single value per key and can multiply rows during a join. Normalize keys or aggregate duplicates upstream when needed.
• Check headers and column positions. The tool creates an output header even for headerless inputs. Setting has_header: "1" on headerless data removes a real data row. Setting "0" on headered data treats the header as a key. With multiple non-key columns per element, old_col_in_header: false repeats that element’s name across its output columns, so use true when distinct labels matter downstream.
• Handle empty elements by need. When an upstream filter can produce empty per-element tables, verify what the join emits for that case and remove the empty elements first if the output must contain only populated sample columns. The PathoGFAIR aggregation workflow filters empty datasets upstream. Other IWC joins do not, so that guard depends on the input and desired output.

For two ordinary tables, tabular-join-on-key covers a direct key join. To stack collection rows into one long table, use tabular-concatenate-collection-to-table. The broader IWC evidence is in iwc-tabular-operations-survey.