Roadmap

A few ideas how to develop the project further, currently a list as a reminder for the main developers themselves, in no particular order, though with a tendency to list more important aspects first:

For version 0.9

  • Documentation

    • Metadata for data acquisition: document info file/refer to publication (and add reference in infofile module documentation)

    • Add “Best Practices” section showing data publications using derived packages (currently only one: Järsvall et al., Chem. Mater. 2022)?

  • Usability

    • Importer/ImporterFactory should issue a warning if no dataset could be loaded, rather than silently continuing, as this often leads to downstream problems and exceptions thrown.

  • Plotting

    • Colorbar for 2D plotter

    • (Arbitrary) lines in plot, e.g. to compare peak positions

      Need to decide whether this goes into plotter properties or gets handled as proper annotations; probably the former, but a good starting point to think about the latter.

  • Processing

    • CombineDatasets: combine data from several datasets into a single dataset; parameters allowing to define the axis values/quantity/unit, possibly even from given metadata; to decide: How to handle metadata that might be invalidated?

    • MetadataUpdate/MetadataChange: Change metadata of a given dataset from within a recipe. Useful in case datasets contain (known) spurious or otherwise inappropriate metadata. (Metadata are provided manually and are therefore prone to human errors).

  • Logging

    • Add loggers from other modules (than task) and derived packages

  • Add export tasks to dataset tasks

  • Recipe-driven data analysis:

    • Better handling of automatically generated filenames for saving plots and reports: unique filenames; using the label rather than the source (id) of the dataset

    • Handling of results: automatically add datasets to dataset list? How to deal with result labels identical to existing datasets?

    • Sub-recipes that can be included in recipes

      Allows for creating a library of recipes for rather complex tasks that can simply be called as single step from another recipe

  • Report task:

    • Operating on recipes, i.e. report on all tasks in a recipe

    • Adding arbitrary dict representations of properties of datasets/results to context

  • Reports:

    • Looking for templates in user directory

  • Documentation:

    • How to debug a recipe?

    • Better document command-line options of the “serve” command

For later versions

  • Convert from collections.OrderedDict to dict, as starting with Python 3.7, dicts preserve the insertion-order of the keys.

  • Plot styles

    • Switch in recipe settings for applying a style to all plots

    • user-defined styles

  • Annotations

    • graphical annotations for characteristic points (and distances, areas?)

  • Remaining basic processing and analysis steps:

    • denoising (via SVD or similar)

    • SNREstimation with explicitly providing noise (using both, processing and analysis)

  • Interpolation

    • for ND with N>2

    • different types of interpolation

  • Templates for creating derived packages

  • Plotter: Factory to create single plots of each given dataset.

  • Basic maths in values of recipes (ranges, basic numpy functions)?

    May impair the platform-independence of the recipe (i.e., tying it to Python/NumPy)

Todos

A list of todos, extracted from the code and documentation itself, and only meant as convenience for the main developers. Ideally, this list will be empty at some point.

Todo

How to handle noisy data in case of area normalisation, as this would probably account for double the noise if simply taking the absolute?

original entry

Todo

  • Make type of interpolation controllable

  • Check for ways to make it work with ND, N>2

original entry

Todo

  • Make type of interpolation controllable

  • Make number of points controllable (in absolute numbers as well as minimum and maximum points with respect to datasets)

original entry

Todo

There is a number of things that are not yet implemented, but highly recommended for a working recipe-driven data analysis that follows good practice for reproducible research. This includes (but may not be limited to):

  • Parser for recipes performing a static analysis of their syntax. Useful particularly for larger datasets and/or longer lists of tasks.

original entry

Todo

Can recipes have LOIs themselves and therefore be retrieved from the extended data safe? Might be a sensible option, although generic (and at the same time unique) LOIs for recipes are much harder to create than LOIs for datasets and alike.

Generally, the concept of a LOI is nothing a recipe needs to know about. But it does know about an ID of any kind. Whether this ID is a (local) path or a LOI doesn’t matter. Somewhere in the ASpecD framework there may exist a resolver (factory) for handling IDs of any kind and eventually retrieving the respective information.

original entry