Tabular: relabel by row counter

Replace each input line with sample_0, sample_1, and so on when downstream needs labels determined solely by row order. tp_awk_tool runs an awk program over the connected dataset. Awk’s NR counts input records starting at one, so subtract one for a zero-based first label. This operation produces a one-column dataset and discards the original row contents.

Use tabular-compute-new-column if original columns must survive. If labels must reflect collection element identifiers, use collection-aware provenance such as tabular-concatenate-collection-to-table. Sorting, filtering, or inserting lines upstream changes these row-order labels.

Whole-row recipe

Connect the input dataset to infile and publish the tool’s outfile. Set code to an awk program and leave variables: [] when no extra awk variables are needed. The IWC exemplar pins toolshed.g2.bx.psu.edu/repos/bgruening/text_processing/tp_awk_tool/9.5+galaxy3.

For a headerless input, the direct zero-based program is:

tool_state:
  code: '{print "sample_" (NR-1)}'
  infile: { __class__: ConnectedValue }
  variables: []

With three input lines it writes sample_0, sample_1, and sample_2, one per line. Use NR without subtraction for sample_1 as the first label. Blank lines also count as records and receive labels. Remove them upstream if they should not count.

The MAGs binning evaluation IWC workflow uses this observed program:

tool_id: toolshed.g2.bx.psu.edu/repos/bgruening/text_processing/tp_awk_tool/9.5+galaxy3
tool_state:
  code: '{gsub( $0 ,"sample_" (NR-1)); print}'
  infile: { __class__: ConnectedValue }
  variables: []

That program calls gsub with the current row as a regular expression, then prints the resulting row. It is evidence of the row-counter idiom, but it is not a general whole-row replacement: regex characters in the input can change what matches, and a row may fail to match itself as a pattern. Use direct print for new whole-row steps.

Headered input

For a one-column input, preserve the first line as a header and give the first data row sample_0:

tool_state:
  code: 'NR==1 {print; next} {print "sample_" (NR-2)}'
  infile: { __class__: ConnectedValue }
  variables: []

For a headered file, NR==2 is the first data row. Using NR-1 in this branch would start it at sample_1. On a multi-column input, preserving the original header would put several header fields above one-column labels. Replace the header with a single field, such as NR==1 {print "sample"; next}, if the output must be a consistent one-column table. If the original columns must remain, use tabular-compute-new-column.

Verification

The local workflow fixture in verification_paths compares zero-based, one-based, and header-preserving output against explicit expected files. It checks the direct-print programs, not the IWC gsub program. The IWC example and tool pin are recorded in iwc-tabular-operations-survey §2g. A successful workflow run should match the expected labels and row count, including the handling of its input header.

See also

• iwc-tabular-operations-survey — §2g and §7 awk split decision.
• tabular-prepend-header — header-preserving variants.
• tabular-compute-new-column — preserve rows while adding/replacing columns.
• tabular-concatenate-collection-to-table — preserve collection element identity instead of synthesizing row labels.