0.2.8

0.2.8 Changelog

• Updated citation information.
• Added specific testing for OSX for Travis, and updated the install method to better pick up issues with pip installations.
• Feature: Added a gallery of pages for each of the datasets included with pyrolite. This will soon be expanded, especially for the reference compositions (to address #38).

pyrolite.geochem

• PR Merged: Kaarel Mänd submitted a pull request to add a number of shale and crustal compositions to the reference database.
• Bugfix: Fixed a bug where lambdas would only be calculated for rows without missing data. Where missing data was present, this would result in an assertion error and hence no returned values.
• Bugfix: Fixed a bug where missing data wasn't handled correctly for calculating lambdas. The functions now correctly ignore the potential contribution of elements which are missing when parameterising REE patterns. Thanks to Steve Barnes for the tip off which led to identifying this issue!
• Feature: Added pyrolite.geochem.ind.REY, pyrolite.geochem.pyrochem.list_REY, and pyrolite.geochem.pyrochem.REY to address
  (#35). This issue was also opened by Kaarel Mänd!
• As a lead-in to a potential change in default parameterisation, you can now provide additional specifications for the calculation of lambdas to
  pyrolite.geochem.pyrochem.lambda_lnREE and pyrolite.util.lambdas.calc_lambdas to determine the basis over which the individual orthogonal polynomials are defined (i.e. which REE are included to define the orthonormality of these functions). For the keyword argument params, (as before) you can pass a list of tuples defining the constants representing the polynomials, but you can now alternatively pass the string "ONeill2016" to explicitly specify the original parameterisation, or :code:"full" to use all REE (including Eu) to define the orthonormality of the component functions (i.e. using params="full"). To determine which elements are used to perform the fit, you can either filter the columns passed to these functions or specifically exclude columns using the exclude keyword argument (e.g. the default remains exclude=["Eu"] which excludes Eu from the fitting process). Note that the default for fitting will remain, but going forward the default for the definition of the polynomial functions will change to use all the REE by default (i.e. change to params="full").
• Significant performance upgrades for pyrolite.geochem.pyrochem.lambda_lnREE and associated functions (up to 3000x for larger datasets).
• Added a minimum number of elements, configurable for pyrolite.geochem.pyrochem.lambda_lnREE. This is currently set to seven elements (about half of the REE), and probably lower than it should be ideally. If for some reason you want to test what lambdas (maybe just one or two) look like with less elements, you can use the min_elements keyword argument.
• Added pyrolite.geochem.pyrochem.list_isotope_ratios and corresponding selector pyrolite.geochem.pyrochem.isotope_ratios to subset isotope ratios.
• Added pyrolite.geochem.pyrochem.parse_chem to translate geochemical columns to a standardised (and pyrolite-recognised) column name format.

pyrolite.plot

• Bugfix: Fixed a bug where arguments processing by :mod:pyrolite.plot.color would consume the 'alpha' parameter if no colour was specified (and as such it would have no effect on the default colors used by matplotlib.pyplot)
• Bugfix: pyrolite.plot.color now better handles colour and value arrays.
• Bugfix: Keyword arguments passed to pyrolite.plot.density will now correctly be forwarded to respective functions for histogram and hexbin methods.
• Customised :mod:matplotlib styling has been added for :mod:pyrolite plotting functions, including legends. This is currently relatively minimal, but could be expanded slightly in the future.
• The bw_method argument for scipy.stats.kde.gaussian_kde can now be parsed by pyrolite density-plot functions (e.g. pyrolite.plot.pyroplot.density, pyrolite.plot.pyroplot.heatscatter). This means you can modify the default bandwidth of the gaussian kernel density plots. Future updates may allow non-Gaussian kernels to also be used for these purposes - keep an eye out!
• You can now specify the y-extent for conditional spider plots to restrict the range over which the plot is generated (and focus the plot to where your data actually is). For this, feed in a (min, max) tuple for the yextent keyword argument.
• The ybins argument for pyrolite.plot.pyroplot.spider and related functions has been updated to bins to be in line with other functions.
• Conditional density pyrolite.plot.pyroplot.REE plots now work as expected, after some fixes for generating reverse-ordered indexes and bins
• Added a filter for ternary density plots to ignore true zeroes.
• Some updates for pyrolite.plot.color for alpha handling and colour arrays .

pyrolite.comp

• Updated transform naming to be consistent between functions and class methods. From this version use capitalised versions for the transform name acronyms (e.g. ILR instead of ilr).
• Added for transform metadata storage within DataFrames for pyrolite.comp.pyrocomp, and functions to access transforms by name.
• Added labelling functions for use with pyrolite.comp.pyrocomp and pyrolite.comp.codata to illustrate the precise relationships depicted by the logratio metrics (specified using the label_mode parameter supplied to each of the resepective pyrolite.comp.pyrocomp` logratio transforms).

pyrolite.util

• Revamped pyrolite.util.classification to remove cross-compatibility bugs with OSX/other systems. This is now much simpler and uses JSON for serialization.
• Small fix for pyrolite.util.plot.style.mappable_from_values to deal with NaN values.
• Added pyrolite.util.log for more streamlined logging (from pyrolite-meltsutil)
• Added pyrolite.util.spatial.levenshtein_distance for comparing sequence differences/distances between 1D iterables (e.g. strings, lists).