CHANGELOG.md

Change Log

All important changes to the decitala package will be documented here.

The changelog format is based on Keep a Changelog and the project uses Semantic Versioning.

v1.1.2 August 17, 2021

Fixed

• Fixed typo for Shattala (formerly Shaltala).
• Added missing Schultz prime contour classes, 12a and 12b, to the contour module.

v1.1.1 August 14, 2021

Fixed

• Rewrote my implementation of the Morris (1993) and Schultz (2008) contour reduction algorithms. Greatly shortened and simplified the code. The Schultz implementation was still complex enough that I decided to move it to a new module (decitala.hm.schultz), but the function schultz_prime_contour is still used in the same way –– it's just wrapped now.
• Added tests for all Rossignol examples from Schultz (2008).

v1.1.0 July 27, 2021

Added

• Added prosodic meters CSV file to the corpora/prosody directory. These encodings map to the new ProsodicMeter object.
• Added exact parameter to FragmentHashTable –– this allows you to more easily create exact-match hash tables for the various corpora.
• Added ditrochee to default_split_dict.

Fixed

• Issues #227: SP plotting utils were incorrectly returning plt.show.

v1.0.0 July 19, 2021

First stable release.

Added

• Added a class method to db.ExtractionData to maintain consistency between it and search.Extraction.
• Added split_dict parameter to search.path_finder. That way one doesn't need to re-write this post-processing code.
• Added a contour_class in hm.contour for experimenting with Schultz's work.
• Added NormalizedCounter and dict_to_ordered_dict to decitala.utils.
• Cleaned up existing hyperparameter calculation and plotting scripts, as well as (re)added a file for hyperparameter optimization (SLSQP and constricted space at finer resolution).
• Added Di- and Tri- proceleusmatic to corpora.
• Switched to codecov.io for coverage; it was annoying to include it as part of my workflow –– I'd rather it be automated.
• Added colors column to ODNC database and corresponding attribute to the Species wrapper.
• Allow the user to choose spearman or pearson correlation in the KS and KS_diatonic functions in hm.hm_utils.
• Added function for getting the "shells" of an iterable to utils: utils.get_shells(iterable).
• Added create_path_database function to database.db for creating a database from fragments in a path.

Changed

• Return type for hm_utils.pc_counter is now either a tone counter or the sum of the quarter length values associated with each pitch class; any normalization now must use hm_utils.normalize_pc_counter.
• The default cost function used in path_finder and the associated helpers is now CostFunction3D with the optimized parameters.
• Renamed create_database to create_extraction_database.
• Renamed poorly named functions in the contour module, e.g. contour.pitch_content_to_contour is now just contour.pitch_contour().

Removed

• Removed unnecessary frag_type parameter to Extraction. Now a class attribute of GeneralFragment (and subclasses).

Fixed

• The return_tonic parameter for KS_diatonic was returning incorrect values.

v0.14.1 July 2, 2021

Fixed

• Fixed a number of typos and added missing examples in decitala.hm.
• The contour_to_prime_contour function still had the include_depth argument. I removed it.
• Re-added commented-out assertions and exceptions, now that all the tests are passing.
• Updated Basic Usage to match the most recent updates.
• Missing default weights to CostFunction3D.

v0.14.0 July 1, 2021

Added

• Moved all moiseaux (private repo) tools to decitala. The decitala.db now holds Transcription and Species SQLAlchemy query wrappers; also added the corresponding helpers: db.get_all_transcriptions() and db.get_all_species().
• Added database.db_utils module.
• Added a make_core_dbs file to decitala.extra that creates the fragment database and ODNC database.
• Added decitala.hm directory for holding all harmony and melody (i.e. pitch) related analysis tools. This directory holds an hm.molt module (tools for dealing with the modes of limited transposition), an hm.contour module for contour calculations (moved some general utils there, like Morris prime contour calculation), and an hm.hm_utils for utility functions; this final module also stores the associated color and key-finding coefficient data.
• Improved implementation of the Morris contour reduction algorithm (1993). Also added an implementation of Schultz's modified contour reduction algorithm (see Schultz 2008, p. 108).
• Added decitala.sp directory for holding signal processing tools. It currently only holds an sp_utils module for spectrogram/audio plotting, but this directory will be populated in future versions.
• Added Dipartite and Tripartite Hexasyllabic Metrics: Dianapest, Dicretic, Didactyl, and Triiamb.
• Scripts for calculating hyperparameters and plotting analysis are in the decitala.extra directory. Latest accuracy is 74.41%.
• Added enforce_earliest_start argument to path_finding_utils.sources_and_sinks.
• Added split() method to both GeneralFragment and Extraction. Used for splitting Di- and Tri- partite greek prosodic feet.
• Added slur_count and slur_start_end_count arguments to the Extraction dataclass.

Changed

• Changed roll_window parameter to window_size (instead of window_length).

Removed

• Dactylo-Epitrite. Removed as, along with Dochmius, there are a large number of variations of the fragment (see Traité Vol. 1).

Fixed

• Graph calculation for Dijkstra was being done within the function. Given the agnostic source/target picking, this was very inefficient; now, the graph is generated a single time in dijkstra_best_source_and_sink.
• Various bugfixes for Dijkstra and Floyd-Warshall for source and target picking.

v0.13.2 June 19, 2021

Fixed

• Added the missing progress bar for path_finding_utils.build_graph (verbose=True did nothing).
• Fixed bug (#144) in measure_divide_mode of utils.get_object_indices.
• Fixed several documentation errors.

v0.13.1 June 19, 2021

Fixed

• Issue #157: broken source button.

v0.13.0 June 19, 2021

Added

• All results from search.rolling_hash_search are now stored as Extraction dataclasses. This new approach greatly simplifies the inconsistent dictionary-style search results. Also removes the need to store frag_types in the hash tables.
• Added a CostFunction class to decitala.path_finding_utils. This cost function class is to be used as an input to path_finder and dijkstra (and floyd_warshall, as well), replacing the previously used weights dictionaries. This addition allows the user to input any cost function they like (see PR#152).
• Added three functions to fragment for querying the fragment database for each fragment type: get_all_greek_feet, get_all_decitalas, and get_all_prosodic_fragments.
• Added rolling_SRR function to utils.
• Added decitala.extra module where miscellaneous scripts will be stored. This will eventually hold the script for computing hyperparameters (decitala.extra.hyperparameters) and the formerly hidden script for remaking analysis files.
• Added decitala dtest --module CLI tool for quickly running doctests (hindered by relative imports). Consequently removed the tests/doctest_runner.py file.
• Restructured the DB approach. Added a database directory that will store all the database-related material.

Changed

• All tree creation has migrated to the treeplotter package. Removed the Treant templates accordingly.
• Greatly simplified the logic in successive_ratio_array and successive_difference_array using numpy.

Fixed

• Issues #147: allow weight hyperparameters as input to Dijkstra; #143: 0 as possible quarter length in FragmentHashTable; #157: broken source button in docs (as well as missing section).

v0.12.1 May 23, 2021

Fixed

• The path_finding_utils.build_graph function was calculating the cost between two vertices without checking if it needed to be calculated.
• Bug in GreekFoot returned incorrect self.data due to extra slash in path.
• Fixed incorrect presets for generate_all_modifications in the hash table module. Also now force clearing the dictionary before running load with new parameters.
• Documentation fixes; added screenshots of the visualization functions.

v0.12.0 May 21, 2021

Added

• Began adding prosodic patterns from Traité T1 analyses. Also added classes (and corresponding frag_types) for the package. You can now create fragment.ProsodicFragment objects as well as create, for instance, a ProsodicFragmentHashTable. These fragments will accumulate over future versions (directory structure subject to change).
• Added plot_2D_search_results to vis module for plotting search results in XY space, along with a given path. (This implements #4, finally!)
• Re-added search.rolling_search_on_array. It conforms to the hash table approach. (See #48)
• Added allow_contiguous_summation parameter to search.rolling_hash_search and path_finder. (See #127)
• Added Breve and Macron classes to the fragments module. This will eventually be used in accessing the parts of a ProsodicFragment.
• Added missing parameter for windows and allow_subdivision in search.path_finder.

Changed

• Big documentation improvements, code/logic cleanup. Trying a new solution for documentation maintenence. Using a github build in the sphinx makefile.
• Removed utils.is_octatonic_collection. All Modes of Limited Transposition stuff will eventually be living in moiseaux.molt.
• Moved some search-specific functions from utils (they were really helpers, not utility functions). See #138.

v0.11.1 May 13, 2021

Changed

• Moved dseg calculation to utils.py that GeneralFragment.dseg now wraps. Also made reduced a parameter in the same function for simplicity (removing the need for a second method).
• Table name in database.py for extractions is now Extractions, not Fragments (this made no sense).
• Structure of documentation. Major improvements to Basic Usage.

Fixed

• Missing lru_cache in GeneralFragment.ql_tuple().
• Missing get_engine/get_session in database.py.
• Issues #133: formatting issue in docs; #131: missing assertions in database tests.

v0.11.0 May 12, 2021

Added

• Added a utils.phrase_divider function for over-simplistically dividing a filepath/part-num combination into phrases (in the same output as utils.get_object_indices). The division is only on the basis of the appearance of rests and fermatas. (See #104)
• Added a contour_to_neume function in utils.py for some experiments with Wai Ling Cheong's work (2008).
• Hash tables and search now support stretch augmentation. Added utils.stretch_augment for simplicity.

Changed

• All functions accepting a ql_array in utils.py now take in a variable called ql_array –– a few were set to fragment instead which was confusing (fixes #122).
• The measure_divider_mode of utils.get_object_indices now accepts "str" and "list" to be consistent with standard python type naming.
• Renamed utils.ts_to_reduced_ts to utils.reframe_ts and now allow any input new_denominator (reducing maximally by default).
• Removed complicated Cauchy-Schwartz pre-filtering in the Tree instantiation. Faster to just overwrite than check.

Fixed

• Issues #130: Allow both lowercase and uppercase algorithm names in search.path_finder; #129: contiguous summation fails with rests; #132: ignore_grace was not doing anything (also fixed docs typo) in get_object_indices;

v0.10.1 May 1, 2021

Changed

• The caching in the fragment module now uses functools.lru_cache(maxsize=None) (replacing the newer functools.cache) to allow support for python 3.7/3.8.

Fixed

• The utils.find_possible_superdivisions function was including itself as a possible superdivision. This was a bit confusing, so I added an include_self parameter. This option was accidentally set to False by default which has been fixed.
• Issues #125: duplicate sr and rsr results from rolling_hash_search; #123/#126: source-sink error in Dijkstra (fixed with a much more general solution (see path_finding.dijkstra.dijkstra_best_source_and_sink) and integrated into search.path_finder.

v0.10.0 April 29, 2021

Added

• Revamped the database module (see #120). Everything is now ported over to SQLAlchemy and is more easily extendable to broader rhythmic corpora. Also added a database.batch_create_database function for creating a database from a large set of compositions.
• Implemented Dijkstra's Algorithm for path finding. This is now the default algorithm used in search.path_finder, but the user can override this.
• New utils functions and additions: a non_retrogradable_measures function for finding all palindromic measures in a given filepath and part number; an optional measure_divider_mode parameter to utils.get_object_indices which returns the same objects, divided into lists of measures or objects divided by a string; a utils.ts_to_reduced_ts function for fully reducing time signatures; a UtilsException class.
• The search.path_finder function now has optional save_filepath argument for dumping the results to a JSON file. (#97)
• All classes inheriting from GeneralFragment (currently just Decitala and GreekFoot) as well as GeneralFragment itself now have a frag_type class attribute.

Changed

• Renamed the fragments directory to corpora. Also removed the .sib encoding files for the greek metrics to make the package lighter weight.
• The fragments module was relying on SQLite in a piecemeal way. Objects from encoded corpora are now proper SQLAchemy models for consistency with the rest of the package.
• The fragment.morris_symmetry_class function now returns integers representing the classes (instead of string describing them). The meaning of each class is given in the documentation.
• Refactored path_finding.floyd_warshall and added helper functions for all path-finding algorithms to path_finding_utils.py.
• The CLI path-finder tool now logs the saved file.
• Improved doctest integration by running them within the (pytest) tests directory.
• Removed all # -*- coding: utf-8 -*- lines.
• Renamed trees.rolling_search to trees.rolling_tree_search to be consistent with the other rolling search type(s) (#114).
• The Decitala.get_by_id class method now requires a string as input. Additionally the id_num property returns a string as a number of the Decitalas have subtalas.
• All GeneralFragment objects now have the carnatic_string and greek_string properties.

Fixed

• Issues #116: missing result logging from decitala.cli.pathfinder; #115: incorrect results from Morris symmetry classes; #73: default value for try_contiguous_summation.
• Incorrect serialization condition for GeneralFragment –– forced incorrect parsing to and from analysis files.

v0.9.1 April 16, 2021

Added

• Added testing to improve coverage (50%-52%)

Fixed

• Issues #102: empty numpy iterator for cases with no extracted fragments; #109: missing verbose argument in search.path_finder; #111: UnboundLocalError for Bach chorale –– this also resolves #101; #106: reformatted analysis JSON files to match the output of search.path_finder (every extraction is a dictionary).
• Typo in Documentation installation instructions.
• Missing documentation for vis.result_bar_plot.
• Fixed incorrect order of Morris Symmetry Classes (1999).

Removed

• Removed py_modules from setup.py.

v0.9.0 April 14, 2021

Added

• Function for annotating a score with the extracted data: vis.annotate_score.
• Added private remake_analysis_files.py for remaking the analysis files in each release.
• Added testing to improve coverage (46%-50%).
• Added various utility functions: utils.measure_by_measure_time_signatures (for for extracting measure-by-measure time signatures); utils.is_octatonic_collection (for checking if pitch content belongs to one of the two octatonic collections); utils.write_analysis (for writing analysis JSON files); utils.net_ql_array (for returning all quarter length values in a given composition as a vector); utils.transform_to_time_scale (for getting the time-scale notation of quarter length array).
• Added PRIMES variable to utils.py for storing primes <100.
• Added vis.result_bar_plot for plotting the counts of the extracted fragments in a path_finder list or analysis JSON (#105).
• Added the first class of the Theorie_Karnatique dataset (Tiśra) to the fragments directory. These are not yet integrated into the fragments module.

Changed

• Improved fragment.py's FragmentEncoder and FragmentDecoder classes with better logic; this fixes #99: incorrect frag_type in the JSON encoding.

Fixed

• Missing import bug in vis.annotate_score.
• README: incorrect installation notes in README, improved translation to the commentary.
• Issues #53: duplicates in logs caused by saved handlers; #100: path_finder bug related to curr_best_sink;

v0.8.4 March 22, 2021

Fixed

• Added the missing Zenodo badge to the README.md file.

v0.8.3 March 22, 2021

Fixed

• Issue #90: removed unused JS modules used in tree visualization.

v0.8.2 March 22, 2021

Added

• Added a .coveragerc file for excluding the tests and CLI tools from coverage reports.

Changed

• Refactored get_path to include a reconstruct_standard_path helper function.

Fixed

• Additional bugfixes for slur constraint in get_path (causing infinite runtime).
• Issues #19: referenced a fault Nary tree size. It appears to be fixed, but I added a test; #95: the pathfinder command line tool now writes to JSON; #94: minor formatting error in floyd_warshall.py.

v0.8.1 March 20, 2021

Added

• Coverage and coverage shield for code improvement.

Changed

• The CLI pathfinder tool now wraps the search.py module's path_finder function.
• The create_tree_diagram now uses webshot from R to save to PNG. Can also now effectively do FragmentTree.show() without opening in the browser using the wand library.

Fixed

• Github Actions is now caching dependencies for faster builds.
• Fixed bugs in vis.create_tree_diagram.
• Bugfixes in path_finder related to is_spanned_by_slur attribute.
• The utils.roll_window now ensures that no NoneTypes are included if the window size is greater than the input data. Similarly, in rolling_hash_search we ensure no searches of size greater than the length of the data.
• Fixed missing Iambs in the GreekFoot hash table (#91).
• The Decitala and GreekFoot classes included a stream attribute in their __init__ –– this is already created in the super.
• Hotfix to extremely strange inheritance bug in the GeneralFragment child classes. (#92)
• Bugfix in floyd_warshall.get_path. Function was inserting the given starting point in the path, even if it was overrided by the slur constraint.

v0.8.0 March 15, 2021

Added

• Added a loader function in utils.py for easy loading of Messiaen's analyses (encoded by me). Also added additional analysis json files (including parts 0 & 1 of Livre d'Orgue (1951-52) movement V).
• Added testing and improved functionality for the hash table creation and rolling_hash_search. This included adding a GreekHashTable and CombinedHashTable for a combined database.
• Added an optional progress bar to the Floyd-Warshall algorithm.
• Added function for getting the best source and sink after Floyd-Warshall (floyd_warshall.best_source_and_sink(data)).
• Added +/- 0.125, +/- 0.375, -0.25, 0.875, 1.75, 2.625, 3.5, 4.375 as possible difference valuess in hash_table.py; also added 0.125, 0.25 as a possible ratio values.
• Added a path_finder command line tool that wraps all the functions for analysis via Floyd-Warshall & the hash tables. Used simply as decitala path-finder --filepath ... --part_num ... --frag_type ....
• Added a ignore_single_anga_class_fragments parameter to rolling_hash_search (default is False).
• Added optional slur_constraint parameter to floyd_warshall.best_path which constrains the path to require the slurred fragments found (#87).

Removed

• Removed the self.conn attribute in the Decitala class; in doing so, we can now use multiprocessing (multiprocessing requires pickling which is impossible on sqlite3 Connection objects).

Fixed

• Missing modification data in rolling_hash_search.

v0.7.4 March 8, 2021 (Kent)

Fixed

• Finished applying flake8 to modules.
• Fixed #83 (missing modules in setup.py)

v0.7.3 March 8, 2021 (Kent)

Fixed

• Re-gitignored local_docs.

v0.7.2 March 8, 2021 (Kent)

Fixed

• Missing modifications types rr, rd, sr, and rsr in DecitalaHashTable. Refactored and clean up the instantiation code –– still requires work for Greek metrics and General fragments.
• Applied flake8 to vis.py.
• JSON Serialization/Deserialization errors for GeneralFragment and its inherited classes. This makes saving "training data" easier. (This step is for the next minor patch.)
• The local_docs directory (this was previously gitignored for no good reason). Also an analyses directory of databases holding valid analysis of compositions.

v0.7.1 March 3, 2021 (NYC)

Fixed

• In enabling github actions, I hit several git/github snags. There were some commit message errors and bad merges. Hopefully everything is fixed now.

v0.7.0 March 3, 2021 (NYC)

Added

• Implemented the Floyd-Warshall Algorithm for path-finding.
• Added a hash_table.py module for more efficient searching. This is used in search.rolling_hash_search.
• Added flake8 for maintaining PEP8 standards. Applied to setup.py, fragment.py, and utils.py.
• Added a fragment_roll function to vis.py (#62).
• Added a progress bar to DBParser.model_full_path to more easily track progress.
• Added a DBParser.onset_ranges method that returns a list of tuples holding the onset ranges of all extracted fragments.
• Added a return_data parameter to model_full_path (just a wrapper for path_data).
• Added an MIT License

Changed

• The trees.py module has been refactored into trees.py and search.py with the relevant search functions in the latter module. (#27)

Fixed

• Issue #54 (unreferenced variable); Issue #51 (updated databases); Issue #58 (unfiltered fragment table);

v0.6.3 February 23, 2021 (NYC)

Added

• Added support/testing for python3.8.

Changed

• Migrated to SQLAlchemy for database.create_database.
• Further refactored database.create_database.
• Default window sizes in database.create_database no longer includes 1.
• Updated the gitignore file.

Fixed

• Issue #45 (requires .db extension in database.create_database); Issue #46 (missing logger reference in helper function); Issue #47 (missing logs in the trees.py module); Issue #40 (incorrect numbers in database.create_database logs); Issue #50 (change start in enumeration statements for readability); Issue #41 (first part of migration to SQLAlchemy); Issue #37 (example SQL insertion error).

v0.6.2 - February 5, 2021 (NYC)

Changed

• Refactored code in database.create_database to improve readability.

v0.6.1 - February 3, 2021 (Kent, CT)

Fixed

• Fixed issue #35. Error with saving logs to file.

Changed

• The trees module no longer logs results. This is a temporary change while I figure out a better solution for global/local logging.

v0.6.0 - February 3, 2021 (Kent, CT)

Added

• Added the missing database.model_full_path function.
• Documentation fixes for DBParser and fragments.py.
• Optional save_logs_to_file parameter to database.create_database.
• Can now create FragmentTrees from a list of multiple data paths.
• Section in the documentation giving a brief exposition to the encoded rhythmic fragments.

Changed

• Documentation now is in sphinx RTD theme.
• Revamped README.
• Rather than setting the same logger at the top of each file, added get_logger function to utils.py in which we can set optional filename to the basicConfig for saving to file when desired.

Fixed

• Missing name argument in fragment trees created from frag_type.

v0.5.2 - January 17, 2020 (Kent, CT)

Removed

• Codecov coverage status (because the service is terrible).

v0.5.1 - January 17, 2020 (Kent, CT)

Added

• Added a function in database.py that filters out cross-corpus duplicates from the data generated in create_database (database.remove_cross_corpus_duplicates).

Changed

• Fragment Trees are now created with frag_types via the class method FragmentTree.from_frag_type. (The logic in the __init__ is now much cleaner.)

Fixed

• Coverage results published on codecov.io (private upload token) with badge.

v0.5.0 - January 16, 2020 (Kent, CT)

Added

• Added a create_fragment_database function in database.py that holds name/ql data & ratio/difference equivalents in the decitala and greek metric databases (including equivalents across databases). This data is available in fragment_database.db in the databases directory.
• Decitala and GreekFoot objects now have an equivalents method that return equivalents (based on the rep_type) in the fragment corpus.
• Created a Travis-CI account for continuous integration.
• Added a Codecov account for coverage.
• Added build icons for travis-ci and codecov.

Changed

• Decitala and Greek Metric instantiation now relies on SQL database instead of os.listdir.
• FragmentTree instantiation now relies on SQL database instead of os.listdir.
• No doctests rely on absolute paths anymore.

v0.4.4 - January 10, 2020 (Kent, CT)

Added

• The output of rolling search now has an id parameter. This will be useful in a number of contexts. NOTE: this may require some fiddling when dealing with combined databases.

Changed

• Rewrote all the pofp.py code, making it far more readable. The duplicate error caused by contiguous summation should now be fixed.

v0.4.3 - January 8, 2020 (Kent, CT)

Changed

• The DBParser now has metadata in the attributes, which stores the path table num, number of subpaths, and average onset data. That way, it needn't be reevalutaed every time we run the model.

Fixed

• Fixed minor documentation errors.

Removed

• Removed all inheritence from object in classes (python3 classes are new-style).

v0.4.2 - January 7, 2020 (Kent, CT)

Added

• Added fragment table visualization (DBParser.show_fragments_table) using pandas (wrapper for pd.read_sql_query). Also added a few other sub-displays (like show_slurred_fragments).
• Added preliminary native FragmentTree visualization. Can be called with FragmentTree.show().
• Implemented first path processing/modeling methods for the database.DBParser class. Most of the paths.py code has moved there.

Changed

• Fragments table from database.create_database has the name in the fragment column instead of full repr.
• Output of get_pareto_optimal_longest_paths now includes all data, not just the fragment.
• Cleaned up the path code in database.create_database.
• The Paths tables in the database now store the row number in the onset_range columns.
• The Path tables are no longer 0-indexed.

Fixed

• The rolling search code now has a line that filters out all grace notes. This was causing the duplicates in the database creation.
• Minor documentation fixes (typos and old parameters).

Removed

• The paths.py file has been removed as all of its functionality has migrated to database.DBParser.

v0.4.1 - January 4, 2020 (Kent, CT)

Fixed

• Bugfix for prime pitch contour calculation (cseg doesn't work on extrema data).
• Bugfix in database creation (missing comma).

v0.4.0 - January 4, 2020 (Kent, CT)

Added

• Added utils.pitch_content_to_contour.
• Implementation of Morris' 1993 pitch contour reduction algorithm. utils.contour_to_prime_contour.
• Added Pitch_Contour column in the Fragment database.

Changed

• The utils.roll_window function now allows for a fn input (allows for rolling window over parts of list, as defined by a lambda expression).

v0.3.2 - January 1, 2020 (Kent, CT)

Added

• __all__ for each module for ease-of-import.
• Documentation for CLI (version getter and cli.create_db).

Fixed

• Bugfix in utils.frame_is_spanned_by_slur. Previously didn't take into account the fact that a range may have multiple overlapping spanners.
• Removed old tree diagram information from cli.

v0.3.1 - December 31, 2020 (Kent, CT)

Added

• Documentation page for vis.create_tree_diagram (Fragment Tree visualization using Treant.js.)
• Log message telling the user if a database has already been made (useful for working in Jupyter).

Fixed

• Incorrect output of pofp.partition_data_by_break_points caused by unsorted data from trees.rolling_search.
• Documentation typo for utils.filter_single_anga_class_fragments.

v0.3.0 - December 30, 2020 (Kent, CT)

Added

• Added preliminary database.DBParser class that allows for easier querying of data from the Fragment table of database.create_database.

Changed

• The pitch information from rolling search is now stored in the Fragments table (made in database.create_database).

Removed

• Helper functions database._check_tuple_in_tuple_range and database._pitch_info_from_onset_range are removed due to the above addition.

v0.2.4 - December 30, 2020 (Kent, CT)

Fixed

• Bugfix in database.create_database (related to new rolling search output).
• Bugfix in pofp.get_pareto_optimal_longest_paths (related to new rolling search output).

v0.2.3 - December 30, 2020 (Kent, CT)

Added

• Added a utils.frame_to_midi option used in rolling search.
• Documentation updates reflecting the new rolling search output.

Changed

• Output of trees.rolling_search is now a list of dict objects. This is better for querying data, adding extra parameters, etc...
• The trees.rolling_search function now stores pitch data using the utils.frame_to_midi.
• Changed the input data parameter of utils.frame_to_ql_array from data to frame (matching the other functions).
• Made the utils.find_clusters function public.

Removed

• Removed trees.rolling_search_on_array (at least for now) as it doesn't currently match the other search formats.
• Removed utils.contiguous_multiplication.

v0.2.2 - December 29, 2020 (Kent, CT)

Added

• Added a .yaml file for a pre-commit hook that prevents writing to master. (Uses pre-commit library).

Changed

• Changed the name of utils.frame_is_spanned_by_slur (fixing a typo).

Removed

• Removed unnecessary doctest imports from modules.
• Removed old cleaning function for setup.py.

v0.2.1 - December 29, 2020 (Kent, CT)

Changed

• The decitala.trees and decitala.database modules now use logging.disable (removing a number of if statements) for readability. (Also removed logging from decitala.pofp.)

v0.2.0 - December 29, 2020 (Kent, CT)

Added

• The decitala.utils module now includes a frame_is_spanned_by_slur function.
• The output of trees.rolling_search now includes is_spanned_by_slur for each fragment found.
• The output of database.create_database now shows which fragments are spanned by music21.spanner.Slur objects.

Fixed

• The database.create_database function now raises a DatabaseException when an invalid score path is provided.

v0.1.1 - December 28, 2020 (Kent, CT)

• First tagged version.
• Documentation (made with sphinx, hosted on github pages) available at https://luke-poeppel.github.io/decitala/.