You're reading an old version of this documentation. For up-to-date information, please have a look at v0.9.

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 0.8

  • Plotting

    • Colorbar for 2D plotter

    • colormaps for multiple lines

  • Processing

    • ExtractSlices (plural): extract several slices from a dataset and combine them in a new dataset

    • 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?

  • 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

  • Processing of 2D (eventually ND with N>1) datasets:

    • Projecting/averaging excluding certain lines (due to artifacts from external noise sources or else)

    • Combining a list of 1D datasets to a 2D dataset (reverse operation of SliceExtraction)

For later versions

  • Get rid of OrderedDict instances, as Python preserves order in dictionaries since version 3.6

  • 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)

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

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

Flesh out these additional DatasetAnnotation classes, particularly in light of the newly created PlotAnnotation classes that may eventually be a way to graphically display the dataset annotations.

original entry

Todo

Clarifly the difference between the HistoryRecord and Record classes, and explain which is used when and how.

original entry

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

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

While generally, recipe-driven data analysis works well in practice, improving usability and robustness is high on the list. This includes ( but may not be limited to) a parser for recipes performing a static analysis of their syntax and is particularly useful for larger datasets and/or longer lists of tasks. Potential steps in this direction:

  • Add check() method to aspecd.tasks.Task

  • Define required parameters in a (private) attribute of either the individual task level or even on the level of the underlying objects

  • Potentially reuse the _sanitise_parameters() method.

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