nmlform

nmlform is a formatter and round-trip parser for Fortran namelists.

Documentation

Unlike value-oriented readers, it keeps the source lines as the source of truth and derives a location-tagged tree over them, so comments, spacing, quoting, and layout survive a round trip. That makes it safe to reformat a namelist, or to edit individual values, without disturbing anything you did not touch.

Installation

nmlform has no runtime dependencies, so a plain virtual environment is all you need:

python -m venv .venv
source .venv/bin/activate  # on Windows: .venv\Scripts\activate

Then install this package into the environment:

git clone https://github.com/ken-lauer/nmlform
cd nmlform
python -m pip install .

To install optional extras, use one of the following:

# Optional parsers for YAML/TOML `--values` files (JSON needs nothing extra):
$ python -m pip install .[all]
# The test suite requirements:
$ python -m pip install .[test]
# The documentation requirements:
$ python -m pip install .[doc]
# Everything:
$ python -m pip install .[all,test,doc]

Command-line usage

Installing the package provides an nmlform console script that reformats one or more namelist files, writing the result to standard output. Use - to read from standard input.

Given a messy example.nml:

&plot a=1  bb = 2.0
    long_name= 'value'   ! a comment
/

Running nmlform reformats it:

$ nmlform example.nml
&plot
  a         = 1
  bb        = 2.0
  long_name = 'value'  ! a comment
/

Common options:

nmlform --in-place *.nml         # rewrite files in place
nmlform --check *.nml            # exit non-zero if any file would change
nmlform --diff *.nml             # show a unified diff of what would change
cat example.nml | nmlform -      # read from stdin, write to stdout

Layout is controlled by flags that mirror the API's formatting options (--indent-size, --tabs, --field-case, --align-equals / --no-align-equals, --align-comments / --no-align-comments, --blank-line-after-group / --no-blank-line-after-group). See nmlform --help for the full list.

Setting and removing values

The nmlform-set console script edits a single namelist file, setting and/or removing values and writing the result to standard output (or to a file with -o, or back to the input file with -i). Edits are surgical: the source layout, comments, and untouched values are preserved verbatim.

Given plot.nml:

&plot
  a = 1  ! a comment
  bb = 2.0
/

$ nmlform-set --set plot a 42 --set plot cc "'new'" --remove plot bb plot.nml
&plot
  a = 42  ! a comment
  cc = 'new'
/

--set and --remove are repeatable. Use NAMELIST#N (1-based) to target the N-th of a repeated group. Pass --reformat to also apply the formatting flags above; otherwise only the edited values change. See nmlform-set --help.

Values can also come from a file with --values (repeatable; - reads stdin):

$ nmlform-set --values overrides.json plot.nml
$ nmlform-set --values overrides.yaml --set plot a 99 plot.nml   # --set wins

The format is inferred from the extension (override with --values-format):

• JSON (built in), YAML, and TOML map groups to {key: value} tables. A string becomes a quoted Fortran string, numbers and booleans become Fortran literals (.true.), a list becomes a value list, and null removes the key. YAML and TOML need the optional parsers: pip install nmlform[all].
• A namelist file (.nml/.init) is applied group by group, field by field, copying each raw value verbatim — handy for merging one namelist's values into another.

For example, overrides.json:

{ "plot": { "a": 42, "title": "hello", "bb": null } }

produces the same edit as --set plot a 42 --set plot title "'hello'" --remove plot bb.

API usage

Parse a file (or a string) into a NamelistFile, then render it. Rendering without options reproduces the source verbatim; passing NamelistFormatOptions reformats it:

from nmlform import NamelistFile, NamelistFormatOptions

nml = NamelistFile.from_file("example.nml")

# Round-trip: byte-for-byte identical to the source.
assert nml.render() == open("example.nml").read()

# Reformat with explicit options.
print(nml.render(NamelistFormatOptions(indent_size=4, field_case="upper")))

Access and edit groups and their assignments. Edits are surgical — only the touched value is spliced, so surrounding comments and layout are preserved:

from nmlform import NamelistFile

nml = NamelistFile.parse("&plot a = 1  bb = 2.0 /\n")

plot = nml.get_namelist("plot")
print(plot.get("a").value)          # -> the value Token for `a`

plot.set("a", "42")                 # update in place
plot.set("cc", "'new'")             # append a new key before the `/`
plot.remove("bb")                   # drop an assignment

print(nml.render())

You can also update or create groups in bulk:

nml.update_namelist("plot", {"a": "1", "title": "'demo'"})

Values are kept as raw source tokens. Use unquote_value / quote_value to convert between a Fortran string literal and its content:

from nmlform import quote_value, unquote_value

quote_value("it's")                 # -> "'it''s'"

See the documentation for the full API reference.