NEXTNANOPY README

nextnanopy is a Python module to interface the nextnano software. This package includes features for:

• Output files: User-friendly method to load the datafiles which allows easy and flexible post-processing.
• Input files: Load input files, set variables, save by finding unused name, execute the file, write input files, etc.
• Configuration: Setup default nextnano configuration (path to executables, databases, licenses, etc).
• Import from GDS files: Load polygons from GDS files and user-friendly methods to generate raw text of nextnano shapes (beta).

Note: A valid license for the nextnano software is not compulsory for the general use of nextnanopy, unless you would like to execute input files via Python.

Future of nextnanopy

Currently, nextnanopy has basic features for nextnano products nextnano++, nextnano³, nextnano.NEGF and nextnano.MSB.

In future releases, we would like to implement:

• Common post-processing methods: Like loading and plotting together bandedges, eigenenergies and eigenvalues, colormap of potential landscape with GDS polygons on top, etc.
• User-friendly input file creation/modification: The idea is to load any input file and it would detect all the different blocks so the user can easily modify parameters like 'boundary conditions' or 'region material'. Similarly, it should be user-friendly to create any input file from scratch via Python.
• Feedback loops: A routine that allows the user to optimize any figure of merit of a device with a given set of input variables. The user can set a post-processing routine to get the figure of merit from the simulated data and later, a new set of input variables will be generated and executed. This feedback loop will repeat until a set of conditions are satisfied.
• Guidelines for contributors: Set of rules if you want to contribute to the project.

Installation

Nextnanopy can be installed on Linux / OS X / Windows. For details, please refer to docs/How to install nextnanopy.md.

Documentation

Basic features are documented in docs/examples. Python scripts in templates/ will help you to start playing with nextnanopy.

Support

Do you want to help nextnanopy? Please send an email to python@nextnano.com.

History of changes

Development version

Version 1.2.0 (July 20th, 2026)

• InputFile(text=...) now builds an input file from a string: the product is detected from the text, the right product class is returned, and the variables are parsed — without reading anything from disk. Previously the text was silently ignored and you got back an empty, product-agnostic object. InputFile(fullpath, text=...) does the same and keeps fullpath as the name to save back to; it used to raise TypeError. Note that InputFile() followed by file.text = ... cannot work and never could — with no text at construction there is nothing to detect the product from, so you get a product-agnostic object whose .variables stays empty. Use InputFile(text=...) instead.
• bug fix: InputFile's parse argument is no longer ignored when passed positionally. InputFile(path, None, True) silently left the file unparsed (.content was None); only the InputFile(path, parse=True) spelling worked. Both now behave the same.
• InputFile cannot be subclassed and now says so with a TypeError. It chooses the class to build from the file's contents, so a subclass was silently discarded and you got a plain product class back. Subclass the product's InputFile (e.g. nextnanopy.nnp.inputs.InputFile), or InputFileTemplate for all products. Related: isinstance(InputFile(path), InputFile) is False — the object you get back is a product class, and InputFileTemplate is the common base to check against and annotate with.
• introduced ExecutionPool (nextnanopy.inputs), a ThreadPoolExecutor-based alternative to ExecutionQueue for running input files concurrently. Not yet used internally (Sweep still uses ExecutionQueue); available for direct use.
• Sweep now has save() and execute() as the preferred, shorter names for save_sweep() and execute_sweep(). For now they simply forward to the existing methods, so nothing changes; in a future release the names will swap (the short names will hold the implementation and the *_sweep names will remain as deprecated aliases). Prefer sweep.save() / sweep.execute() in new code.
• convergence check no longer hangs headless runs: check_convergence(mode='pause') asks how to proceed only when an interactive terminal is attached; without one (CI, cluster jobs) it now raises like mode='terminate' instead of blocking on input.
• nextnanopy.config.set(...) now reaches every input file you create afterwards, with no save() needed — configure first, then build your input files. Previously a set() was ignored unless it was followed by save(). The configuration is bound when the input file is created: it takes a copy of nextnanopy.config as it stands at that moment, so later changes to nextnanopy.config do not reach files that already exist, and editing file.config changes neither nextnanopy.config nor any other input file. The input files a Sweep generates follow the same rule. Passing configpath still takes precedence over all of this.
• importing nextnanopy no longer creates or checks ~/.nextnanopy-config. The config is built on first access of nextnanopy.config (or when an InputFile is created without configpath).
• saving the config file is now atomic: an interrupted write can no longer leave a half-written .nextnanopy-config behind.
• Config with an empty path now raises ValueError; a path to a non-existing file yields an empty configuration that can be filled in and saved.
• bug fix: ExecutionQueue no longer busy-spins at 100% CPU while waiting for parallel simulations (limit_parallel > 1). It now sleeps between checks (ExecutionQueue.poll_interval, default 0.1 s) and fills all free parallel slots at once; wall time is unchanged.
• performance: creating an InputFile reads the file (and sets up the config) only once; product detection loads the file once instead of once per product; DataFile loads the data from disk only once (product-specific .txt files still need two reads).
• DataFile restructured: the per-product DataFile classes are removed, only per-product .txt loaders remain. Loader autodetection now raises NotImplementedError with a clear message when no loader fits.
• removed nextnanopy.defaults.get_DataFile, which mapped a product name to its DataFile class. Use nextnanopy.DataFile(path, product=...) instead.
• execute is no longer exported at the top level: nextnanopy.execute is gone. It was the low-level command runner, which requires exe, license, database and outputdirectory to be passed by hand — the very things nextnanopy reads from your configuration for you. Use InputFile.execute() (as in all the examples), or import nextnanopy.commands.execute directly if you really want the raw call.
• nextnanopy.commands.execute now raises a clear ValueError for an empty or non-existing input file path before starting a simulation.
• bug fix: nested or concurrent loops over the same DictList, InputFile or DataFile no longer interfere with each other. Iterating the same object in an inner and an outer loop (e.g. for v1 in datafile: for v2 in datafile: ...) used to end the outer loop early, because the iteration position was stored on the object itself. Iteration still yields the values (Variable/Coord objects, input variables), so no API change.
• bug fix: the detected product for invalid input files is consistently lowercase 'not valid'
• bug fix: replaced deprecated xml usage in nextnano.NEGF_classic input handling
• removed the unused is_{product}_input_file functions (superseded by is_{product}_input_text)
• internal: savetxt in utils/misc refactored; first unit tests for ExecutionQueue added and sped up; nextnano++ syntax updated in the test example input file

Version 1.1.1 (July 13th, 2026)

• bugfixes
• enforce ruff linting and formatting

Version 1.1.0 (May 5th, 2026)

• most of the templates are moved to the nextnano tools distribution packages, no longer part of this repo
• new parameter parse for InputFile (default False): content parsing of nextnano++ input files is now opt-in. Use InputFile(fullpath, parse=True) to populate file.content with the block structure. With the default parse=False the file loads normally — variables are available, content is None — which also allows loading files where preprocessor directives cause unbalanced {} that the parser cannot handle.
• bugfixes

Version 1.0.5 (Mar 24th, 2026)

• bugfix

Version 1.0.4 (Mar 24th, 2026)

• saving sweeps in temporary location: Sweep.save_sweep(temp=True)
• bugfix with json output file of the sweep
• templates updates

Version 1.0.3 (Jan 8th, 2026)

• now modern MSB input files are supported
• use dandelion instead of golden in plotting pallette
• enforce alphabetical order in DataFolder tree

Version 1.0.2 (Oct 10th, 2025)

• bugfix: now supports the avs binary files with ascii coords
• added nextnano mpl style for plotting, see utils
• added possibility to save input files in temporary folder: InputFile.save(temp=True)
• improved gds import

Version 1.0.1 (Jul 12th, 2025)

• .txt files are read as Dat without coords (all columns are variables)

Version 1.0.0 (Jul 11th, 2025)

• support of python3.13
• NEGF renamed to NEGF_classic (requires .nextnanopy-config update)
• NEGF++ renamed to NEGF (requires .nextnanopy-config update)

Version 0.1.22 (May 7th, 2024)

• improved Dat loader

Version 0.1.21 (Feb 28th, 2024)

• Added DataFile loader for total_charges.txt
• nextnanopy now supports python3.11

Version 0.1.20 (Dec 13th, 2023)

• nextnanopy is aligned to be used with nextnanoevo. nextnanoevo will be released in 2024

Version 0.1.19 (Jul 21st, 2023)

• new feature: saving DataFile: DataFile.save
• new feature: DataFolder.find_multiple - search files by multiple keywords
• new property DataFolder.name - return the basename of the folder

Version 0.1.18 (Jul 04th, 2023)

• new product available: nextnano.NEGF++
• setting the text for InputFile is available again (but switching product when setting text os forbidden)
• interface with .gds files is extended

Version 0.1.17 (Jun 07th, 2023)

• fixed a bug with nnp.shapes.GdsPolygons.get_obelisks
• get_obelisks can be used again. The usage is deprecated, nnp.shapes.GdsPolygons.get_polygonal_prisms is prefered way to convert gds to nextnano++ structure

Version 0.1.16 (Apr 25th, 2023)

• nextnanopy now supports python 3.10!
• The execute_sweep function has been updated to include a new parameter, separate_sweep_dir, which allows the user to specify whether the output files should be saved in a separate directory or a common output directory. By default, execute_sweep will still create a separate output directory for each sweep. However, if the user sets separate_sweep_dir to False, the output files will be saved in a common output directory instead.
• ListDict now supports negative indexes (in the same way as a normal python list)
• nnpp assistant is updated
• shapely is no longer necessary to import geometry from gdsii file

Version 0.1.15 (Aug 31st, 2022)

• input files now can be executed in parallel through nextnanopy.inputs.ExecutionQueue class
• Sweep uses ExecutionQueue and executes several simulations in parallel if parallel_limit>1 in Sweep.execute_sweep(parallel_limit = n)

Version 0.1.14 (Jul 27th, 2022)

• nextnanopy.Sweep.save_sweep has parameter round_decimal to limit round of digits in the folder names. Deafault round_decimal = 8

Version 0.1.13 (Jun 13th, 2022)

• Default .dat loader of nextnanopy.DataFile was changed. Before x,y,z and position were coordinates, anything else recognized as variables. Now first column of the '.dat' file is recognized as coordinate, all other columns are variables.
• Demo: parser of the whole input file for nextnano++. Raise error if file contains incorrect number and order of parentheses '{}'

Version 0.1.12 (Mar 14th, 2022)

• added support of input variables for nextnanoNEGF. Note: after save all non-xml data (i.e. comments) is not preserved.

Version 0.1.11 (Jan 10th, 2022)

• Feature: DataFile.plot() - make a preview plot. Graph for 1-dimensional data and a colormap for 2-dimensional data.
• Feature: New class DataFolder. Designed to simplify navigation in output directory.
• Beta feature: postprocess.simple_optimize. Runs simulation over given variables space and optimizes the desired output value.

Version 0.1.10 (Oct 06th, 2021)

• Feature: Added new class Sweep allowing to create automatic sweeps over few variables

Version 0.1.9 (Dec 16th, 2020)

• Feature: InputFile.execute has now show_log (True or False) to turn on/off the log. Note: the log file is always saved.
• Removed default messages when saving an input file

Version 0.1.8 (Dec 14th, 2020)

• Buf fix: find unused name when saving input file

Version 0.1.7 (Dec 3rd, 2020)

• docstring for inputs.py, outputs.py, config.py, mycollections.py, datasets.py
• Updated examples in docs/examples/

Version 0.1.6 (Dec 2nd, 2020)

• Feature: [DataFile] access with index to coordinates and to variables

df = nextnanopy.DataFile(...)
df['x'] # same as df.coords['x']
df['Gamma'] # same as df.variables['Gamma']

• Feature: [DataFile] is loopable as well as .coords and .variables
• Feature: [InputFile] access with index to input variables

df = nextnanopy.InputFile(...)
df['xmax'] # same as df.variables['xmax']

• Feature: [InputFile] is loopable as well as .coords and .variables
• Feature: [InputFile] extra attributes

df = nextnanopy.InputFile(...)
df.folder_output # returns the output folder if it was executed, otherwise it raises an error
df.filename      # settable
df.filename_only # settable
df.folder_input  # settable

• Feature: User-friendly information for DataFile, InputFile, Variable, Coord and InputVariable
• Bug fix: .vtr dataset reshape method

Version 0.1.5 (Dec 2nd, 2020)

• Feature: Support of .vtr data files
• Feature: If unit is not found, default value is 'a.u.'.
• Bug fix: Better methods to find correctly name and unit from data file headers

Version 0.1.4 (Nov 29th, 2020)

• Feature: Create an empty input file and set the raw text with the attribute text.

Version 0.1.3 (Nov 28th, 2020)

• Bug fix: Find unused name when save input files.
• Feature: Default label for Variables and Coords. The attribute label returns "name (unit)".

Version 0.1.2 (Nov 20th, 2020)

• Bug fix: Raw text of input variables without comment was generated incorrectly.

Version 0.1.1 (Nov 18th, 2020)

• Bug fix: Find the home path for OSX and Linux.

Version 0.1.0 (Nov 13th, 2020)

• Initial release