Version 1.1 update; JOSS draft

Version updated to 1.1.  A draft paper for JOSS submission has been
added to the `papers` directory.

f90nml/__init__.py

@@ -6,7 +6,7 @@
 from f90nml.namelist import Namelist
 from f90nml.parser import Parser
 
-__version__ = '1.0.2'
+__version__ = '1.1'
 
 
 def read(nml_path):

papers/joss/paper.md

@@ -0,0 +1,88 @@
+---
+title: 'f90nml - A Python module for Fortran namelists'
+tags:
+    - fortran
+    - namelist
+    - python
+authors:
+    - name: Marshall L. Ward
+      orcid: 0000-0002-4944-7431
+      affiliations: '1, 2'
+affiliations:
+    - name: NOAA Geophysical Fluid Dynamics Laboratory, Princeton, NJ, USA
+      index: 1
+    - name: National Computational Infrastructure, Canberra, Australia
+      index: 2
+date: 22 May 2019
+---
+
+# Summary
+
+F90nml is a Python module used for importing, manipulating, and writing Fortran
+namelist files.  The primary use case is to read a namelist variable into a
+`Namelist` data structure, which behaves similarly to the intrinsic `dict`
+type.  The content can be read and modified using standard Python methods, and
+can be saved as a formatted namelist file.
+
+Basic features include support for all intrinsic data types, as well as
+user-defined types and multidimensional arrays.  User-defined types are
+interpreted as hierarchical tree of `Namelists`\s.  Multidimensional arrays
+are saved as nested lists of lists, with the most innermost lists corresponding
+to the first dimensional index.  This reverses the index order in Python, but
+corresponds to the usual ordering in memory.
+
+F90nml also includes a `patch` feature, which allows one to modify the values
+of an existing namelist while retaining its comments or existing formatting.
+There is some limited ability to add or remove values when patching.
+
+Because data type is set at runtime in the executable and is not specified in
+the namelist, the data type of each value must be inferred, usually based on
+the strictest interpretation of the value.  Weak typing rules within namelists,
+such as the optional use of string delimiters or the rules for interpreting
+logical values, can lead to further ambiguity, including across compilers, and
+various control flags are provided to manage such behavior.  In most cases, an
+ambiguous value will fall back to interpretation as a string.
+
+Another limitation of the namelist format is the use of a arbitrary start index
+in a Fortran array, which may be used at runtime but not specified in the
+namelist.  For this reason, arrays are assumed to begin at the lowest explicit
+index which exists in the namelist, which is stored as metadata.  For example,
+if we parse the namelist below:
+
+    &a_nml
+       x(3:4) = 1.0, 1.1
+       x(6:7) = 1.2, 1.3
+    /
+
+then it would be saved internally as the following 0-based Python list:
+
+    nml = {
+        'a_nml': {
+            'x': [1.0, 1.1, None, 1.2, 1.3]
+         }
+         '_start_index': {'x': 3}
+    }
+
+When index is unspecified, then the index is left unspecified within the
+`Namelist`, although ordering remains 0-based within the Python environment.
+Additional control flags are also provided to control the start index.
+
+F90nml also provides the following additional features:
+
+- A command line tool for working in a shell environment
+
+- Lossless conversion between `Namelist` and `dict` types
+
+- Support for legacy Fortran namelist formats
+
+- Conversion between JSON and YAML output
+
+- Automated, configurable output formatting
+
+The module is supported by an extensive test suite with a very high level of
+code coverage, ensuring compatibility of existing namelists over future
+releases.
+
+This project is sustained by the feedback of its users, and continues to
+benefit from contributions from its userbase, for which the author is immensely
+grateful.