ms3 1.2.0: New extraction features

ms3 requires Python 3.10

Extraction of all lyrics

This version enables the extraction of lyrics with all verses. Previously, only the last verse's syllable for any given position was extracted. The lyrics now can be found in [lyrics_[verse]]{.title-ref} columns in the chords facet, where [lyrics_1]{.title-ref} corresponds to the first or only verse.

Extraction of figured bass

Figured bass labels can now be found in the chords facet tables. Score that include at least one figure will have a thoroughbass_duration column and each layer of figures comes in a separate thoroughbass_layer_# column. For example, if all stacks of figures have only layer, there will be only the column thoroughbass_layer_1.

Extraction of custom-named spanners

Spanners with adjusted "Begin text" property get their own columns in the chords tables, containing the relevant subselection of IDs. For example, if a score contains normal 8va spanners and others where the "Begin text" has been set to custom, all IDs will be combined in the column Ottava:8va as before, but the subset pertaining to the custom spanners is additionally shown in the column Ottava:8va_custom.

Including and excluding paths

It is now possible to specify specific directories to be included or excluded from a view, not only folder names.

New methods and properties

• Parse.get_facet() (singular)
• Corpus.fnames
• Corpus.add_dir()
• first version of utils.merge_chords_and_notes()

ms3 1.1.2

ms3 requires Python 3.10

• Refines the new "writing score headers" functionality and makes it non-default. User needs to
  set ms3 metadata --prelims which replaces the flag --ignore that had been introduced in 1.1.1.
• A couple of bug fixes, including a very important one regarding conversion of fifths introduced with b0ce8a1

ms3 1.1.1

• enables updating score headers from the respective metadata.tsv columns
  (title_text, subtitle_text, composer_text, lyricist_text, and part_name_text)
• Parse, Corpus, and Piece now come with the method keys()

ms3 1.1.0: Experimental MuseScore 4 support

This version does not throw errors when trying to parse files created by MuseScore 4. Parsing these files has not
sufficiently been tested but so far it was looking good. The fact that MuseScore 3 is able to read such files shows
that not much has changed in the file format itself.

The command ms3 convert has been updated to support MuseScore 4 executables. With the current MuseScore 4.0.0 this
is not quite straightforward because conversion to .mscz via the commandline isn't currently working and conversion
to .mscx, if it works at all, deletes the contents of the target directory
(issue #15367). The new function
utils.convert_to_ms4() offers a workaround that creates temporary directories to store the
"Uncompressed MuseScore folder" and then copies the .mscx file to the target directory (default) or zips the
temporary directory into an .mscz file (parameter --format mscz). For all other target formats, the output
will correspond to what the MuseScore 4 executable yields.

ms3 1.0.4

ms3 requires Python 3.10

ms3 has gotten a makeover and does not quite like it did before. The major changes are:

• The library is now optimized for one particular folder structure, namely [meta-corpus ->] corpus -> piece.
• ms3 now comes with a full-fledged "views" feature which lets you subselect files in manifold ways.
• The TSV outputs have gained additional columns. In particular, all TSV files now come with the column quarterbeats reflecting each event's offset from the piece's beginning.
• Warnings concerning irregularities, e.g. wrong measure numbering due to a cadenza, can now be sanctioned by copying them into an IGNORED_WARNINGS file.

New features

• Each object that the user interacts with, Parse, Corpus, and Piece, comes with at least two views, called "default" and "all". The "default" view disregards review files, scores in convertible formats, and scores that are not listed in the top-level metadata.tsv file.
• metadata.tsv files, by the virtue of their first column fname, now serve as authority on what is included in the corpus and what belongs together. This column is always unique and supposed to be used as index.
• Suffixed metadata_<suffix>.tsv files are loaded as available views based on the column fname (other columns are disregarded).
• The Parse object now detects if the passed directory contains individual corpora or if it is a corpus itself.
• Parse objects perform operations by iterating over Corpus objects.
• Corpus objects perform operations by iterating over Piece objects.
• Corpus objects reflect exactly one folder, the corpus_path, and always discover all present files (which can be filtered before the actual parsing). Default output paths are derived from it.
• Piece objects unite the various files pertaining to the same fname and are able to keep multiple versions of the same type apart (e.g., scores or annotation files) and pick one automatically, if necessary, or ask for user input.
• The command ms3 review combines the functionalities of ms3 check, ms3 extract, and ms3 compare, and is now the only command used in the new dcml_corpus_workflow action. For each score that has DCML harmony labels, it stores another score and TSV file with the suffix _reviewed in the folder reviewed.
  • The score has all out-of-label tones colored in red and
  • the TSV file contains a report on this coloring procedure. Both files are stored in the folder reviewed on the top level of the corpus.
  • (1.0.2) In addition, if any warnings pop up, they are stored in the top-level warnings.log file.
• Inserting labels into scores is accomplished using the new method load_facet_into_scores() which comes with the optional parameter git_revision which allows loading TSVs from a specific commit.
• Therefore, ms3 compare (and hence, ms3 review) is now able to compare the labels in a score with those in a TSV file from an older git revision.
• ms3 extract -F extracts form labels and expands them into a tree-like view in the output TSV.

Changes to the interface

• Many things have been renamed for the benefit of a more homogeneous user interface.
  • Methods previously beginning with output_ were renamed to store_.
  • Parse.parse_mscx() => Parse.parse_scores()
• The properties for retrieving DataFrames from Score objects:
  • are now methods and accept the parameters unfold and interval_index.
  • return None when a facet is not available.
• Parsed scores and dataframes are always returned with File object that identifies the parsed file in question. This is particularly relevant when using the get_facet() methods that may return facets from parsed TSV files or extract them from the scores, according to availability.
• Gets rid of the argument simulate except for writing files.
• logger_cfg now as **kwargs
• (1.0.3) Currently the -d/--dir argument to ms3 commands accepts only one directory, not several.

Changes to the outputs

• (1.0.1) When unfolding repeats, add the column mn_playthrough with disambiguated measure Numbers ('1a', '12b', etc.).
• The column label_type has been replaced and disambiguated into harmony_layer (0-3, text, Roman numeral, Nashville, guitar chord) and regex_match (containing the name of the regular expression that matched first).
• Notes tables now come with the two additional columns name (e.g. "E#4") and octave. For unpitched instruments, such as drumset, the column name displays the designated instrument name (which the user can modify in MuseScore), and have no value in the octave columns.
• For pieces that don't have first and second endings, the TSVs come without a volta column.
• Extracted metadata
  • (1.0.1) come with the new columns last_mc_unfolded, last_mn_unfolded, volta_mcs, guitar_chord_count, form_label_count, ms3_version, has_drumset
  • uses the column fname as index
  • comes with a modified column order
  • renames the previous column rel_paths to subdir, whereas the new column rel_path contains
  • include the text fields included in a score. Columns are composer_text, title_text, subtitle_text, lyricist_text, part_name_text.
• Upon a full parse (i.e. if the view has default settings), each facet folder gets a csv-metadata.json file following the CSVW standard. This file indicates the version of ms3 that was used to extract the facets. The version is also included in the last row of the README.

Other changes

Many, many bugs have died on the way. Also:

• Most functions and methods now come with type hints.
• New unittest suite that makes use of the DCMLab/unittest_metacorpus repo and enforces it to be at the correct commit.
• The parser is now more robust against user-induced strangeness in MuseScore files.
• (1.0.1) Repetitions are unfolded for checking the integrity of DCML phrase annotations in order to deal with voltas correctly.
• (1.0.3) Pedal notes that have multiple (volta) endings, although still not being correctly propagated into each ending,
  get propagated into the first ending, and don't cause propagation nor the integrity check to fail anymore

ms3 1.0.3

ms3 requires Python 3.10

ms3 has gotten a makeover and does not quite like it did before. The major changes are:

• The library is now optimized for one particular folder structure, namely [meta-corpus ->] corpus -> piece.
• ms3 now comes with a full-fledged "views" feature which lets you subselect files in manifold ways.
• The TSV outputs have gained additional columns. In particular, all TSV files now come with the column quarterbeats reflecting each event's offset from the piece's beginning.
• Warnings concerning irregularities, e.g. wrong measure numbering due to a cadenza, can now be sanctioned by copying them into an IGNORED_WARNINGS file.

New features

• Each object that the user interacts with, Parse, Corpus, and Piece, comes with at least two views, called "default" and "all". The "default" view disregards review files, scores in convertible formats, and scores that are not listed in the top-level metadata.tsv file.
• metadata.tsv files, by the virtue of their first column fname, now serve as authority on what is included in the corpus and what belongs together. This column is always unique and supposed to be used as index.
• Suffixed metadata_<suffix>.tsv files are loaded as available views based on the column fname (other columns are disregarded).
• The Parse object now detects if the passed directory contains individual corpora or if it is a corpus itself.
• Parse objects perform operations by iterating over Corpus objects.
• Corpus objects perform operations by iterating over Piece objects.
• Corpus objects reflect exactly one folder, the corpus_path, and always discover all present files (which can be filtered before the actual parsing). Default output paths are derived from it.
• Piece objects unite the various files pertaining to the same fname and are able to keep multiple versions of the same type apart (e.g., scores or annotation files) and pick one automatically, if necessary, or ask for user input.
• The command ms3 review combines the functionalities of ms3 check, ms3 extract, and ms3 compare, and is now the only command used in the new dcml_corpus_workflow action. For each score that has DCML harmony labels, it stores another score and TSV file with the suffix _reviewed in the folder reviewed.
  • The score has all out-of-label tones colored in red and
  • the TSV file contains a report on this coloring procedure. Both files are stored in the folder reviewed on the top level of the corpus.
  • (1.0.2) In addition, if any warnings pop up, they are stored in the top-level warnings.log file.
• Inserting labels into scores is accomplished using the new method load_facet_into_scores() which comes with the optional parameter git_revision which allows loading TSVs from a specific commit.
• Therefore, ms3 compare (and hence, ms3 review) is now able to compare the labels in a score with those in a TSV file from an older git revision.
• ms3 extract -F extracts form labels and expands them into a tree-like view in the output TSV.

Changes to the interface

• Many things have been renamed for the benefit of a more homogeneous user interface.
  • Methods previously beginning with output_ were renamed to store_.
  • Parse.parse_mscx() => Parse.parse_scores()
• The properties for retrieving DataFrames from Score objects:
  • are now methods and accept the parameters unfold and interval_index.
  • return None when a facet is not available.
• Parsed scores and dataframes are always returned with File object that identifies the parsed file in question. This is particularly relevant when using the get_facet() methods that may return facets from parsed TSV files or extract them from the scores, according to availability.
• Gets rid of the argument simulate except for writing files.
• logger_cfg now as **kwargs
• (1.0.3) Currently the -d/--dir argument to ms3 commands accepts only one directory, not several.

Changes to the outputs

• (1.0.1) When unfolding repeats, add the column mn_playthrough with disambiguated measure Numbers ('1a', '12b', etc.).
• The column label_type has been replaced and disambiguated into harmony_layer (0-3, text, Roman numeral, Nashville, guitar chord) and regex_match (containing the name of the regular expression that matched first).
• Notes tables now come with the two additional columns name (e.g. "E#4") and octave. For unpitched instruments, such as drumset, the column name displays the designated instrument name (which the user can modify in MuseScore), and have no value in the octave columns.
• For pieces that don't have first and second endings, the TSVs come without a volta column.
• Extracted metadata
  • (1.0.1) come with the new columns last_mc_unfolded, last_mn_unfolded, volta_mcs, guitar_chord_count, form_label_count, ms3_version, has_drumset
  • uses the column fname as index
  • comes with a modified column order
  • renames the previous column rel_paths to subdir, whereas the new column rel_path contains
  • include the text fields included in a score. Columns are composer_text, title_text, subtitle_text, lyricist_text, part_name_text.
• Upon a full parse (i.e. if the view has default settings), each facet folder gets a csv-metadata.json file following the CSVW standard. This file indicates the version of ms3 that was used to extract the facets. The version is also included in the last row of the README.

Other changes

Many, many bugs have died on the way. Also:

• Most functions and methods now come with type hints.
• New unittest suite that makes use of the DCMLab/unittest_metacorpus repo and enforces it to be at the correct commit.
• The parser is now more robust against user-induced strangeness in MuseScore files.
• (1.0.1) Repetitions are unfolded for checking the integrity of DCML phrase annotations in order to deal with voltas correctly.
• (1.0.3) Pedal notes that have multiple (volta) endings, although still not being correctly propagated into each ending,
  get propagated into the first ending, and don't cause propagation nor the integrity check to fail anymore

ms3 1.0.2

ms3 requires Python 3.10

ms3 has gotten a makeover and does not quite like it did before. The major changes are:

• The library is now optimized for one particular folder structure, namely [meta-corpus ->] corpus -> piece.
• ms3 now comes with a full-fledged "views" feature which lets you subselect files in manifold ways.
• The TSV outputs have gained additional columns. In particular, all TSV files now come with the column quarterbeats reflecting each event's offset from the piece's beginning.
• Warnings concerning irregularities, e.g. wrong measure numbering due to a cadenza, can now be sanctioned by copying them into an IGNORED_WARNINGS file.

New features

• Each object that the user interacts with, Parse, Corpus, and Piece, comes with at least two views, called "default" and "all". The "default" view disregards review files, scores in convertible formats, and scores that are not listed in the top-level metadata.tsv file.
• metadata.tsv files, by the virtue of their first column fname, now serve as authority on what is included in the corpus and what belongs together. This column is always unique and supposed to be used as index.
• Suffixed metadata_<suffix>.tsv files are loaded as available views based on the column fname (other columns are disregarded).
• The Parse object now detects if the passed directory contains individual corpora or if it is a corpus itself.
• Parse objects perform operations by iterating over Corpus objects.
• Corpus objects perform operations by iterating over Piece objects.
• Corpus objects reflect exactly one folder, the corpus_path, and always discover all present files (which can be filtered before the actual parsing). Default output paths are derived from it.
• Piece objects unite the various files pertaining to the same fname and are able to keep multiple versions of the same type apart (e.g., scores or annotation files) and pick one automatically, if necessary, or ask for user input.
• The command ms3 review combines the functionalities of ms3 check, ms3 extract, and ms3 compare, and is now the only command used in the new dcml_corpus_workflow action. For each score that has DCML harmony labels, it stores another score and TSV file with the suffix _reviewed in the folder reviewed.
  • The score has all out-of-label tones colored in red and
  • the TSV file contains a report on this coloring procedure. Both files are stored in the folder reviewed on the top level of the corpus.
  • (1.0.2) In addition, if any warnings pop up, they are stored in the top-level warnings.log file.
• Inserting labels into scores is accomplished using the new method load_facet_into_scores() which comes with the optional parameter git_revision which allows loading TSVs from a specific commit.
• Therefore, ms3 compare (and hence, ms3 review) is now able to compare the labels in a score with those in a TSV file from an older git revision.
• ms3 extract -F extracts form labels and expands them into a tree-like view in the output TSV.

Changes to the interface

• Many things have been renamed for the benefit of a more homogeneous user interface.
  • Methods previously beginning with output_ were renamed to store_.
  • Parse.parse_mscx() => Parse.parse_scores()
• The properties for retrieving DataFrames from Score objects:
  • are now methods and accept the parameters unfold and interval_index.
  • return None when a facet is not available.
• Parsed scores and dataframes are always returned with File object that identifies the parsed file in question. This is particularly relevant when using the get_facet() methods that may return facets from parsed TSV files or extract them from the scores, according to availability.
• Gets rid of the argument simulate except for writing files.
• logger_cfg now as **kwargs

Changes to the outputs

• (1.0.1) When unfolding repeats, add the column mn_playthrough with disambiguated measure Numbers ('1a', '12b', etc.).
• The column label_type has been replaced and disambiguated into harmony_layer (0-3, text, Roman numeral, Nashville, guitar chord) and regex_match (containing the name of the regular expression that matched first).
• Notes tables now come with the two additional columns name (e.g. "E#4") and octave. For unpitched instruments, such as drumset, the column name displays the designated instrument name (which the user can modify in MuseScore), and have no value in the octave columns.
• For pieces that don't have first and second endings, the TSVs come without a volta column.
• Extracted metadata
  • (1.0.1) come with the new columns last_mc_unfolded, last_mn_unfolded, volta_mcs, guitar_chord_count, form_label_count, ms3_version, has_drumset
  • uses the column fname as index
  • comes with a modified column order
  • renames the previous column rel_paths to subdir, whereas the new column rel_path contains
  • include the text fields included in a score. Columns are composer_text, title_text, subtitle_text, lyricist_text, part_name_text.
• Upon a full parse (i.e. if the view has default settings), each facet folder gets a csv-metadata.json file following the CSVW standard. This file indicates the version of ms3 that was used to extract the facets. The version is also included in the last row of the README.

Other changes

• (1.0.1) Checks integrity of DCML phrase annotations after unfolding repeats to correctly deal with voltas.
• Most functions and methods now come with type hints.
• New unittest suite that makes use of the DCMLab/unittest_metacorpus repo and enforces it to be at the correct commit.
• The parser is now more robust against user-induced strangeness in MuseScore files.

ms3 1.0.1

ms3 requires Python 3.10

ms3 has gotten a makeover and does not quite like it did before. The major changes are:

• The library is now optimized for one particular folder structure, namely [meta-corpus ->] corpus -> piece.
• ms3 now comes with a full-fledged "views" feature which lets you subselect files in manifold ways.
• The TSV outputs have gained additional columns. In particular, all TSV files now come with the column quarterbeats reflecting each event's offset from the piece's beginning.
• Warnings concerning irregularities, e.g. wrong measure numbering due to a cadenza, can now be sanctioned by copying them into an IGNORED_WARNINGS file.

New features

• Each object that the user interacts with, Parse, Corpus, and Piece, comes with at least two views, called "default" and "all". The "default" view disregards review files, scores in convertible formats, and scores that are not listed in the top-level metadata.tsv file.
• metadata.tsv files, by the virtue of their first column fname, now serve as authority on what is included in the corpus and what belongs together. This column is always unique and supposed to be used as index.
• Suffixed metadata_<suffix>.tsv files are loaded as available views based on the column fname (other columns are disregarded).
• The Parse object now detects if the passed directory contains individual corpora or if it is a corpus itself.
• Parse objects perform operations by iterating over Corpus objects.
• Corpus objects perform operations by iterating over Piece objects.
• Corpus objects reflect exactly one folder, the corpus_path, and always discover all present files (which can be filtered before the actual parsing). Default output paths are derived from it.
• Piece objects unite the various files pertaining to the same fname and are able to keep multiple versions of the same type apart (e.g., scores or annotation files) and pick one automatically, if necessary, or ask for user input.
• The command ms3 review combines the functionalities of ms3 check, ms3 extract, and ms3 compare, and is now the only command used in the new dcml_corpus_workflow action. For each score that has DCML harmony labels, it stores another score and TSV file with the suffix _reviewed. The score has all out-of-label tones colored in red and the TSV file contains a report on this coloring procedure. Both files are stored in the folder reviewed on the top level of the corpus.
• Inserting labels into scores is accomplished using the new method load_facet_into_scores() which comes with the optional parameter git_revision which allows loading TSVs from a specific commit.
• Therefore, ms3 compare (and hence, ms3 review) is now able to compare the labels in a score with those in a TSV file from an older git revision.
• ms3 extract -F extracts form labels and expands them into a tree-like view in the output TSV.

Changes to the interface

• Many things have been renamed for the benefit of a more homogeneous user interface.
  • Methods previously beginning with output_ were renamed to store_.
  • Parse.parse_mscx() => Parse.parse_scores()
• The properties for retrieving DataFrames from Score objects:
  • are now methods and accept the parameters unfold and interval_index.
  • return None when a facet is not available.
• Parsed scores and dataframes are always returned with File object that identifies the parsed file in question. This is particularly relevant when using the get_facet() methods that may return facets from parsed TSV files or extract them from the scores, according to availability.
• Gets rid of the argument simulate except for writing files.
• logger_cfg now as **kwargs

Changes to the outputs

• (1.0.1) When unfolding repeats, add the column mn_playthrough with disambiguated measure Numbers ('1a', '12b', etc.).
• The column label_type has been replaced and disambiguated into harmony_layer (0-3, text, Roman numeral, Nashville, guitar chord) and regex_match (containing the name of the regular expression that matched first).
• Notes tables now come with the two additional columns name (e.g. "E#4") and octave. For unpitched instruments, such as drumset, the column name displays the designated instrument name (which the user can modify in MuseScore), and have no value in the octave columns.
• For pieces that don't have first and second endings, the TSVs come without a volta column.
• Extracted metadata
  • (1.0.1) come with the new columns last_mc_unfolded, last_mn_unfolded, volta_mcs, guitar_chord_count, form_label_count, ms3_version, has_drumset
  • uses the column fname as index
  • comes with a modified column order
  • renames the previous column rel_paths to subdir, whereas the new column rel_path contains
  • include the text fields included in a score. Columns are composer_text, title_text, subtitle_text, lyricist_text, part_name_text.
• Upon a full parse (i.e. if the view has default settings), each facet folder gets a csv-metadata.json file following the CSVW standard. This file indicates the version of ms3 that was used to extract the facets. The version is also included in the last row of the README.

Other changes

• (1.0.1) Checks integrity of DCML phrase annotations after unfolding repeats to correctly deal with voltas.
• Most functions and methods now come with type hints.
• New unittest suite that makes use of the DCMLab/unittest_metacorpus repo and enforces it to be at the correct commit.
• The parser is now more robust against user-induced strangeness in MuseScore files.

ms3 1.0.0

ms3 has gotten a makeover and does not quite like it did before. The
major changes are:

• The library is now optimized for one particular folder structure, namely [meta-corpus ->] corpus -> piece.
• ms3 now comes with a full-fledged "views" feature which lets you subselect files in manifold ways.
• The TSV outputs have gained additional columns. In particular, TSV files now come with the column quarterbeats reflecting each event's offset from the piece's beginning.
• Warnings concerning irregularities, e.g. wrong measure numbering due to a cadenza, can now be sanctioned by copying them into an IGNORED_WARNINGS file.

New features

• Each object that the user interacts with, Parse, Corpus, and Piece, comes with at least two views, called "default" and "all". The "default" view disregards review files, scores in convertible formats, and scores that are not listed in the top-level metadata.tsv file.
• metadata.tsv files, by the virtue of their first column fname, now serve as authority on what is included in the corpus and what belongs together. This column is always unique and supposed to be used as index.
• Suffixed metadata_<suffix>.tsv files are loaded as available views based on the column fname (other columns are disregarded).
• The Parse object now detects if the passed directory contains individual corpora or if it is a corpus itself.
• Parse objects perform operations by iterating over Corpus objects.
• Corpus objects perform operations by iterating over Piece objects.
• Corpus objects reflect exactly one folder, the corpus_path, and always discover all present files (which can be filtered before the actual parsing). Default output paths are derived from it.
• Piece objects unite the various files pertaining to the same fname and are able to keep multiple versions of the same type apart (e.g., scores or annotation files) and pick one automatically, if necessary, or asks for user input.
• The new command ms3 review combines the functionalities of ms3 check, ms3 extract, and ms3 compare, and is now the only command used in the new dcml_corpus_workflow action. For each score that has DCML harmony labels, it stores another score and TSV file with the suffix _reviewed. The score has all out-of-label tones colored in red and the TSV file contains a report on this coloring procedure. Both files are stored in the folder reviewed on the top level of the corpus.
• Inserting labels into scores is accomplished using the new method load_facet_into_scores() which comes with the optional parameter git_revision which allows loading TSVs from a specific commit.
• Therefore, ms3 compare (and hence, ms3 review) is now able to compare the labels in a score with those in a TSV file from an older git revision.
• ms3 extract -F extracts form labels and expands them into a tree-like view in the output TSV.

Changes to the interface

• Many things have been renamed for the benefit of a more homogeneous user interface.
  • Methods previously beginning with output_ were renamed to store_.
  • Parse.parse_mscx() => Parse.parse_scores()
  • Parse.parsed_mscx => Parse.n_parsed_scores
  • Parse.parsed_tsv => Parse.n_parsed_tsvs
• The properties for retrieving DataFrames from Score objects
  • are now methods and accept the parameters unfold and interval_index.
  • return None when a facet is not available.
• Parsed scores and dataframes are always returned with a File object that identifies the parsed file in question. This is particularly relevant when using the get_facet() methods that may return facets from parsed TSV files or extract them from the scores, according to availability.
• Gets rid of the argument simulate except for writing files.
• logger_cfg now as **kwargs

Changes to the outputs

• The column label_type has been replaced and disambiguated into harmony_layer (0-3, Jazz chord, Roman numeral, Nashville, tex) and regex_match (containing the name of the regEx that matched first).
• Notes tables now come with the two additional columns name (e.g. "E#4") and octave. For unpitched instruments, such as drumset, the column name displays the designated instrument name (which, znfortznately, the user can modify in MuseScore), and have no value in the octave columns.
• For pieces that don't have first and second endings, the TSVs come without a volta column.
• Extracted metadata
  . - uses the column fname as index
  • comes with a modified column order
  • renames the previous column rel_paths to subdir, whereas the new column rel_path contains folder + file name of the relevant score
  • include the text fields included in a score. Columns are composer_text, title_text, subtitle_text, lyricist_text, part_name_text.
• Upon a full parse (i.e. if the view has default settings), each facet folder gets a csv-metadata.json file following the CSVW standard. This file indicates the version of ms3 that was used to extract the facets. The version is also included in the last row of the README.

Other changes

• Most functions and methods now come with type hints.
• New unittest suite that makes use of the DCMLab/unittest_metacorpus repo and enforces it to be at the correct commit.
• The parser is now more robust against user-induced strangeness in MuseScore files.

Minor improvements

• recognizes metadata fields reviewers and annotators also in their singular forms
• adds column n_onset_positions to metadata.tsv
• interval index levels renamed from 'iv' => 'interval'
• gets rid of pandas deprecation warnings
• bug fixes & log messages