Tabular: split into collection by key

Use split_file_to_collection in tabular → by column mode when a table must become a list collection for a mapped Galaxy tool. The selected column supplies the element identifiers. Rows with the same resulting identifier go into the same dataset, so a one-row-per-element result requires unique identifiers after any regex replacement.

The SRA manifest workflow cuts its input to an accession column, splits it into a collection, then connects that collection to fasterq_dump for map-over. The variation-reporting workflow uses the same tabular column mode on a combined results table. Both pin split_file_to_collection/0.5.2.

Choose the grouping key and header

For a table with one header line, the following input produces two collection elements:

sample  value
alpha   2
beta    5
alpha   3

The display uses spaces for alignment. The input file is tab-separated. With id_col: "1", top: "1", and the unchanged identifier regex, element alpha contains the header and both alpha rows. Element beta contains the header and its one row. The verification workflow checks the element IDs and complete files. Its Galaxy/Planemo test passed on release_25.1. The behavior was also checked by running the upstream splitter script on the example.

This is the relevant Galaxy-exported tool state from the SRA workflow, with unrelated fields omitted. The input connection targets split_parms|input, and the tabular collection output is list_output_tab.

tool_id: toolshed.g2.bx.psu.edu/repos/bgruening/split_file_to_collection/split_file_to_collection/0.5.2
tool_state:
  split_parms:
    select_ftype: tabular
    input: { __class__: ConnectedValue }
    top: "1"
    split_by:
      select_split_by: col
      id_col: "1"
      match_regex: (.*)
      sub_regex: \1

id_col is one-based. top copies that many leading lines into every element and excludes them from grouping. Set it to 0 for a headerless table. match_regex and sub_regex transform the selected column value into the element identifier. The wrapper’s column-splitting test exercises both header copying and identifier replacement. The cited wrapper checkout declares version 0.5.3, while the IWC workflows above pin 0.5.2.

Check the collection before map-over

• Confirm the chosen column identifies the intended work unit. Repeated values combine rows into one element. If the next tool expects one accession or sample per invocation, check that each key is unique after regex replacement.
• Inspect the resulting identifiers. Replacing different values with the same name combines their rows, which can silently change the mapped job count.
• Check whether the downstream tool expects a header in every element. The splitter copies top lines without interpreting them, and top: "0" leaves no copied header.
• Check a representative element’s contents, not only the number of elements. The split preserves whole input rows, including the identifier column.

For a table already represented as a collection, continue with collection operations. To combine compatible collection elements into one table, use tabular-concatenate-collection-to-table. For the prepare → map → reshape sequence, see manifest-to-mapped-collection-lifecycle.