awk (Compute)

When To Use

Use the awk compute for parsing jobs that simpler computes cannot express: reshaping free-form command output into semicolon-separated rows, correlating multi-line records, or computing derived values. The script receives the source's current result as its input — the raw text when available (e.g. commandLine output), otherwise the table serialized as semicolon-separated lines — and its printed output becomes the new table.

MetricsHub embeds JAWK, a Java implementation of the AWK language, extended with a few MetricsHub utility functions. Standard AWK constructs work; do not assume GNU awk (gawk) extensions are available.

This compute transforms an existing source's result. If the script itself is the source (computing over an explicit input), use the awk source^[1] instead.

Syntax

sources:
  volumes:
    type: commandLine
    commandLine: cmd /c ${file::diskpart.bat}
    computes:
    # The script prints: MSHW;ID;Label;Letter;VolumeType;FileSystem;Size;Status
    - type: awk
      script: ${file::diskPart.awk}
      keep: ^MSHW;
      separators: ;
      selectColumns: "2,3,4,5,6,7,8"

script may also be written inline as a YAML block scalar (see LinuxService for a real example).

Properties

Table Transformation Example

Suppose diskPart.awk prints one MSHW;-prefixed line per volume; keep, separators, and selectColumns then clean up its output:

Between the two, the script's serialized output looked like:

MSHW;0;System;C;Partition;NTFS;254568116160;Healthy
MSHW;1;Data;D;Partition;NTFS;999784378368;Healthy

The keep: ^MSHW; filter drops any stray script output, and selectColumns removes the MSHW marker column.

Recommended Pattern

• Prefix intentional output lines with a marker (MSHW; by convention) and filter with keep: ^MSHW; — this is the standard community pattern (DiskPart, lmsensors, LinuxNetwork).
• Emit semicolon-separated values with a constant column count so mapping stays predictable.
• Keep long scripts in dedicated .awk embedded files referenced with ${file::...}; reserve inline block scalars for short scripts.
• Put the awk compute first in the pipeline, then refine with cheaper computes (translate, keepOnlyMatchingLines…).

Common Mistakes

• Relying on GNU awk extensions (gensub, asort, coprocesses…) that JAWK does not provide.
• Setting selectColumns without separators: the lines are never split, so the selection does nothing.
• Expecting keep/exclude to filter the script's input — they filter its output. Filter input inside the script itself.
• A script that prints nothing empties the table, and every downstream compute and mapping silently gets no rows.

Community Examples

• DiskPart^[2]
• lmsensors^[3]
• LinuxService^[4] (inline script)

From DiskPart, included directly from the connector source:

  # Run Source(1) through an AWK script
  # ID;Label;Letter;VolumeType;FileSystem;Size;Status;MoreInformation
- type: awk
  script: ${file::diskPart.awk}
  separators: ;
  keep: ^MSHW;
  selectColumns: 2,3,4,5,6,7,8,9