Changes.md

What's new in Incanter

Changes for 1.9.0

This is first preview of the Incanter 2.0. This release incorporates the code implemented as part of the GSoC 2014 project "Incanter and core.matrix integration" done by Aleksandr Sorokoumov, aka Gerrrr. (For full list of changes made during GSoC, please look to wiki).

Please, note that this release includes only part of changes planned for Incanter 2.0

Main changes

• Incanter now uses only core.matrix functions with :vectorz as default implementations. You can use another implementation by adding corresponding dependency to your project file & selecting implementation using the set-current-implementation function.
• Incanter's dataset is now based on the Dataset from core.matrix that supports both labeled columns & rows. Dataset is also a matrix now. Matrix functions work on datasets;
• Many functions are now deprecated in favor of corresponding functions from core.matrix. These functions have :deprecated flag in meta-information, for full list please look to wiki.

Breaking changes

• incanter.core/matrix called on seq will produce vector object instead of row-matrix;
• Matrices are not treated as sequences anymore, so calling first, rest, map, reduce, filter, etc on matrices is not generally supported (although might still work on some implementations, e.g. clatrix). It's recommended to use corresponding functions from core.matrix;
• API of the functions decomp-cholesky, decomp-svd, decomp-eigenvalue, decomp-lu, decomp-qr is changed to match core.matrix's implementation. See core.matrix wiki for more details;
• incanter.core/dataset API is changed. Now it accepts column names and seq of rows, matrix, map of columns or seq of row maps as arguments;
• For datasets, the consecutive integers (.e.g 0, 1, 2, ...) are now used as default column names;
• incanter.core/sel does not support automatic cast between strings and keywords anymore;
• 2 arguments version of incanter.core/col-names removed. Please use clojure.core.matrix.dataset/select-columns instead.

Improvements

• Instead of using scripts to perform task in every subdirectory, you can now use lein-sub & lein-modules plugins to do the same. Just execute lein sub test or lein modules install to perform testing, or installing of every Incanter's module;
• Clojure 1.6.0 is used.

Bugfixes

• kendalls-tau now returns correct value.

Known issues

• incanter.stats/sample-wishart throws error during execution;
• For Clatrix-based implementations, it's not always possible to construct matrix from result of map/filter/...

Changes for 1.5.6

Bugfixes

• median returns Double/NaN when it receives empty list (PR #263);
• sel returns dataset when :rows :all specified, independent on size/structure of the dataset (PR #259);
• Fixed calculation of kurtosis (PR #260);
• Fixed plots legend on group-by when repeated rows for the legend column are present (PR #253);
• Fix for problems with gamma distribution (Issue #245);
• Fix for rank-index function (PR #261, #262).

Enhancements

• New functions in incanter.charts to work with series (PR #278):
  • has-series? - checks, does the chart have the series with given name;
  • remove-series - removes given series from chart;
  • extend-line - adds new data to existing series, or creates new series if it doesn't exist.
• save-svg can also accept the OutputStream, not only the file name (PR #279).

Changes for 1.5.5

Bugfixes

• Fix the repl scripts (issue #215)
• Correctly handle from-repo parameter of get-dataset function
• Correctly scale parameter in sample-model-params
• Columns are explicitely casted to str in read-dataset
• Correct different problems found by Eastwood lint tool
• Correctly handle missing values when using log-axis (issue #210)
• BFGS minimization routine uses gradient-fn to estimate f-prime when it is not provided. This also entailed changing gradient-fn to accept matrices as well as vectors.
• xy-plot and time-series-plot modified to take into account group-by names correctly for legend labels (issue #216)
• draw for uniform-distribution uses specified min & max parameters
• Correlation coefficient now 0.0 (not NaN) with constant vector
• Fix for linear model adjusted R-squared (issue #194)
• Catch divide by zero exceptions when calculating rho-k in fmin-bfgs
• Correctly handle :legend for time-series-plot
• Fix for minus for a single argument (issue #195)
• chisq-test works correctly when :x is one sample collection

Enhancements

• New functions in incanter.core:
  • aggregate performs the aggregation of data on given fields (issue #223)
  • get-column-id returns keyword version of column-key if convenient
• New functions in incanter.stats:
  • implementation of gamma-coefficient function
  • concordant-pairs function
• read-dataset now accepts the :comment-char parameter that specifies the commentary character

Breaking changes

• (issue #245) Incanter used the 'rate' as parameter name, although in reality this was a 'scale' parameter (https://en.wikipedia.org/wiki/Gamma_distribution). For pdf-gamma, cdf-gamma & sample-gamma functions the new parameter :scale was introduced (equivalent to the old :rate parameter), and :rate parameter is now the 1/:scale...

Changes for 1.5.4

Made an error during deployment of 1.5.3, so 1.5.4 was released to fix this problem. Don't use 1.5.3!

Changes for 1.5.3

Bugfixes

• Issue 183: the pow & atan2 functions weren't implemented for matrices & datasets.
• read-dataset now converts empty fields to nil, or user-supplied value (see Issue 182).
• Documentation improvements.

Enhancements

• New functions in incanter.charts:
  • set-point-size to control size of points on scatter plots.
• New functions in incanter.core:
  • rename-cols allows to rename columns of dataset
  • replace-column replaces data in column of dataset with new values
  • add-column allows easier to add new column to dataset
  • add-derived-column adds a column to a dataset that is a function of existing columns
  • melt implements part of functionality of R's melt function from reshape package.

Changes for 1.5.2

Bugfixes

• Issue 168: the view function wasn't defined for Matrix class
• Issue 161: maximal idx for slider wasn't correctly calculated
• The sel function on nil was implemented, preventing from getting errors when there was no data specified in the $data variable
• Issue 169: metadata wasn't added to ncol & nrow functions
• Issue 164: to-vect was implemented only for Matrix class, now it works with any support data type
• Issue 165: sel returns a dataset even if result has one row, and we're selecting columns.

Enhancements

• Issue 166: You can use logarithmic axes (with different bases) in Incanter charts. See issue for more details

Changes for 1.5.1

Bugfixes

• Issue 157: when transform-with was used with Matrix, then source data was modified instead of working on copy of data
• Issue 160: when 2-arguments version of solve was used, the exception was thrown

Changes for 1.5.0

Enhancements

• incanter-core's matrix uses native BLAS through jBLAS/Clatrix - this greatly improves performance (on 64-bit Linux see "Known issues" section).
• several interpolation functions were added to incanter-core module (as incanter.interpolation namespace).
• a new option is added to heat-map - :include-zero?.
• the tail function was added.
• new function reorder-columns for a dataset that changes the order of appearance of the datset columns. It does not alter the row order.
• save will print data to standard output if "-" is specified as file name.
• sel and other functions ($, head, tail, etc.) can be used with lists (java.util.List).
• the toeplitz function was added to generate Toeplitz matrix for given vector.
• the scatter-plot-matrix function was added to incanter.chart module.
• incanter.optimize was extended with minimize and maximise functions for performing unconstrained nonlinear optimization using the BFGS algorithm.
• the new incanter.svg module provides save-svg function to output charts to SVG files. Include incanter-svg as dependency to use this functionality.

Many bugfixes

• for function & parametric plots, line is finished in max-range point.
• permutation matrix is returned in LU decomposition.
• linear-model now correctly calculates t-probs.
• fixed division by zero in linear-model for some data.

Modified behavior

• sel will return dataset when :rows or :cols are non-numbers - this changes previous behaviour when list was returned if only one row or col was specified.
• mult & mmult always return matrices, even if it's 1x1 matrix.

Known issues

• conj'ing of matrix & vector doesn't work with new Clatrix - you can either use bind-rows, or wrap vector into another vector:

  (def M (matrix [[0 1] [2 3] [4 5]])) (conj M [6 7]) ; => doesn't work (bind-rows M [6 7]) ; => works (conj M [[6 7]]) ; => works

• decomp-qr performs only full QR decomposition, and the :type parameter is ignored.

• On 64-bit Linux you need to install libgfortran3 package. See jblas wiki for more details

Updated dependencies

• Clatrix: 0.3.0
• Clojure: 1.5.1
• JLine: 2.11

Changes for 1.4.x

Major changes are:

• The incanter.sql module was added to allow load datasets from databases using ClojureQL.
• Support for parametric plots in incanter.chart module
• Bugfixes

Changes for 1.3.x

Major changes are:

• Switch to Leiningen 2 for development
• Incanter-processing was removed. It's recommended to use Quil instead
• Updated versions of dependencies:
  • Clojure: 1.4.0
  • Parallelcolt: 0.10.0
  • math.combinatorics: 0.0.3
  • Apache POI (for incanter.excel): 3.8
  • clj-time: 0.4.4
  • Congomongo: 0.3.3
• incanter.core/get-input-stream & incanter.core/get-input-reader are removed in favor of input-stream & reader from clojure.java.io
• Many bugfixes -- thank you for all people who sent us pull requests!