Skip to content

Tune the YAML formatting

Every file kubesplit writes is formatted by yamkix. The same yamkix styling options are exposed on the kubesplit CLI, so you can control exactly what the generated YAML looks like.

For quote handling specifically, see Control quotes. For the full option list, see the CLI options reference.

Document start and end markers

  • Explicit start (---) is on by default. Disable it with -n/--no-explicit-start.
  • Explicit end (...) is off by default. Enable it with -e/--explicit-end.
# no leading '---', add a trailing '...'
kubesplit --input all-in-one.yml --output out --no-explicit-start --explicit-end

apiVersion: v1
kind: Namespace
metadata:
  name: demo
...

Sequence indentation (dash inwards)

By default, dashes in sequences are pushed inwards (indented under their key). Use -d/--no-dash-inwards to keep them at the sequence level.

Default (dash_inwards on):

spec:
  containers:
    - name: web
      image: nginx:1.27

With --no-dash-inwards:

spec:
  containers:
  - name: web
    image: nginx:1.27

Flow style vs block style

kubesplit emits block style by default. Enable YAML flow style (JSON-like {}/[]) with -f/--default-flow-style:

kubesplit --input all-in-one.yml --output out --default-flow-style

Note

In the default rt (round-trip) parsing mode, ruamel.yaml remembers the style of each collection it parsed, so --default-flow-style mostly affects collections that carry no such memory. Existing block-style content generally stays block style.

Force block style everywhere (--enforce-block-style)

Because rt mode remembers the original style, an input that already contains inline (flow-style) collections keeps them. Use -B/--enforce-block-style to rewrite every flow-style map/list into block style, regardless of how it was written:

kubesplit --input all-in-one.yml --output out --enforce-block-style

Input with an inline data map:

data: {key1: value1, key2: value2}

Output with --enforce-block-style:

data:
  key1: value1
  key2: value2

Note

--enforce-block-style overrides --default-flow-style (everything ends up block style). Empty collections ([] and {}) are kept as-is, since block style has no representation for them. This option only applies in the default rt parsing mode.

Line width

The maximum line width defaults to 2048 (effectively "don't wrap"). Lower it with -w/--line-width if you want yamkix to fold long lines:

kubesplit --input all-in-one.yml --output out --line-width 80

Comment spacing

By default, end-of-line comments are left where they are. Use -s/--spaces-before-comment to normalize the number of spaces between content and a trailing comment:

kubesplit --input all-in-one.yml --output out --spaces-before-comment 2

apiVersion: extensions/v1beta1  # exactly two spaces before the comment

Warning

-s is --spaces-before-comment in kubesplit (an integer), not a "stdout" flag. Kubesplit has no STDOUT mode — see Split descriptors.

Align end-of-line comments

Use -a/--align-comments to line up trailing (end-of-line) comments within each map/list to a common column:

kubesplit --input all-in-one.yml --output out --align-comments

Input:

data:
  short: a # first
  a_longer_key: b # second

Output with --align-comments:

data:
  short: a        # first
  a_longer_key: b # second

Comments are aligned to the widest line within the same map or list, so nested blocks are aligned independently. Combine it with -s/--spaces-before-comment to also control the gap before short comments.

Parser mode: keep or drop comments

The parser mode is rt (round-trip) by default, which preserves comments. Switch to -t safe/--typ safe to drop all comments:

kubesplit --input all-in-one.yml --output out --typ safe

Next steps