Change Log

Version 1.0.36 - Jan 21st, 2024

After numerous issues with peoples old (2020-2021 sklearn issues, have required 1.3 onwards in setup.py file now. ) This meant that had to change python to be 3.7+, as skitlearn only supports 3.8-3.11. People with older installations can pip install earlier versions

Version 1.0.31 to Version 1.0.34 - Dec, 2023

Fixing additional issues related to pandas2 and shift to csv file paths.

Version 1.0.31 - November 17th, 2023

Sorted issues with .pkl incompatability for calibration datasets by resaving all as csv files and loading those.

Version 1.0.31 - November 17th, 2023

Sorted issues with .pkl incompatability for calibration datasets by resaving all as csv files and loading those.

Version 1.0.30 - October 6th, 2023

Issue with Amp-Liq melt mathcing, particularly if non unique values for liqiud names. Fixed so dont end up with a duplicate column name that errors out.

Version 1.0.28 - October 6th, 2023

Saved pkl files for Jorgenson and Petrelli were failing with Sklearn 1.3. retrained models using this version + released new version on github for Thermobar_onnx. Sklearn also changed ‘mse’ to now be called ‘squared_error’ For Cpx_all functions, as doesnt do voting anyway, swapped to onnx versions, so at least those work when people have wrong versions.

Fixed bug with Jorgenson - was doing normalizatoin using the Sample_ID_Liq_num column for MonteCarlo simulations, resulted in liquid getting less and less data as you go to higher iterations. Added if statement to drop column before normalization stage.

Version 1.0.25 - July 6th, 2023

Added CHOMPI from Blundy (2022) function calculate_CHOMPI

Version 1.0.20 Mayn 30th, 2023

Cpx matching function thros error if one column all Nans, have done a check if column there first. Thanks Divya!

Version 1.0.19 April 1st, 2023

Raises warning if people enter non equal length dataframes into functions which dont melt match. Add melt match for plag-fspar (Thanks Jordan!)

Version 1.0.18 April 1st, 2023

Support for non numerical sample names for Opx-Cpx, Opx-Liq matching, Fspar-Liq, Amp-Liq.

Version 1.0.16 April 1st, 2023

Added ability to do confidence intervals in matplotlib

Version 1.0.14 March 13th, 2023

Felix boschetty showed that viscosity doesnt work if sample name is a string. Have changed to drop the sample column during the norm step.

Conversion between fo2 and melt redox only used high T QFM value. Have now changed to allocate based on entered T_K

Version 1.0.12, Feb 13th, 2023

If a user entered Kspar comps into plag-liq functions, returned Nan for eq tests, broke loop. Have fixed to add warning if you enter An<0.05, and also to not return yes or no if T is Nan Have also updated examples to get users to plot input fspar on a ternary diagram.

Version 1.0.11, Feb 11th, 2023

Fixed bug, in old version, T<1300 K, code returned ‘Pass’ for An-Ab in plag-liq, even if fail. Kd value was correct, just string was wrong. Thanks to Bryant platt for spotting this one.

Version 0.12 Nov 23rd, 2021

Added in capability for Plag-Liq, Kspar-Liq temperature calculations for melt matching, and Plag-Liq hygrometers.

Added in warning if users specify Suffix, but already have the suffix.

Version 0.11 Nov 15th, 2021

Fixed issue with indent on Opx-Liq melt matching.

Previous changes before detailed change log started:

Removed Wang et al. (2021) barometer eq. 3 and thermometer eq 2, updated eq 1 and 2 to reflect those in the final version of the manuscript.

Changed order of Cpx-Liq oxides when training Petrelli (2020) to get perfect benchmark