* Compatibility with conda version of geng; remove geng tool from package

* Incorporate pkl files into connectivity database

* Add nauty as dependency

* Add pickle as test dependency

* Switch from strings to pickles for connectivity graphs

* Use blob instead of text to store pickled dictionary

* No longer write substructures to .smi

* Add option to build to select only frequent substructures

* Add connectivity filter to k_configs

* Incorporate connectivity filter into MSn build method

* Build substructures for each set of masses independently

* Call itertools.product on substructures within multiprocessing portion of build

* Configure run script for current create_isomorphism_database inputs

* Built subsets should be empty list, not None

* Update variable names, remove debug options, update docstrings

* Add annotate_msn and generate_structures user functions

* Move stage at which multiprocessing step is performed

* Allow for multiple output options in build

* Remove ppm option for retrieving elemental composition from substructure db

* Allow list of mc/exact_mass to be passed to generate_structures

* Use TemporaryDirectory to store unittest results

* Let generate_structures return/yield smiles

* Implement build_msn to incorporate considerations for building structures from MS/MS

* Implement annotate_msn to provide an interface to build_msn

* Add/update build docstrings

* Remove unnecessary build parameters

* Pass data dictionary to user-facing build functions rather than separate mc, exact_mass, MSn masses

* Update variable naming conventions

* Add newline between smiles in out file

* Update SubstructureDb for removal of .pkl files

* Add function create_substructure_database

* Bring tests up to date with variable renaming

* Bring scripts up to date with variable renaming

* Simplify loading of test data and remove teardown

* Remove unused class ConnectivityDb and update SubstructureDb parameters

* Implement additional non-msn build tests

* Improve temporary table cleaning logic

* Fix issues with new build functions

* Allow tests to load auxiliary test data

* Implement msn tests and update k_config test for new parameter

* Correctly specify ppm in generate_structures

* Minor docstring and code reformatting

* Add binder dir

* Add example notebook

* Remove scripts

* Implement basic notebook

* Add small substructures to database prior to msn annotation

* Complete notebook example

* Fix logic for when smi_out_dir is None

* Rename example_msms.ipynb to workflow.ipynb

* Add pip to install metaboblend

* Add data dir, remove databases dir, move test data to data dir

* Write notebook databases to notebook_data

* Unzip test data

* Simplify test paths

* Remove databases from gitignore

* Use test databases for notebook

* Implement simple hydrogenation rules

* Get bond types rather than number of available atoms for hydrogen rule calculations

* Don't count dummy atoms for bond type calculations

* Remove dummy atom mass

* Use max_degree of 6 and 2 available_atoms by default for create_substructure_database

* Account for the fact we use neutral peaks (i.e. have removed adduct ion)

* Modify hydrogen re-arrangement rules for doulbe bonds

* Update databases tests

* Implement test for calculate_possible_hydrogenations using reference numbers

* Add test for calculate_hydrogen_rearrangements

* Update hydrogen re-arrangement calculation function documentation

* Update remaining unit tests

* Add hydrogen re-arrangement compound HMDB XMLs

* Record even substructures

* Record even substructures in results DB

* Add indexes to improve combine_ecs function performance

* Improve results DB hierarchy and implement aggregation of scoring metrics

* Define SQLite functions to calculate scores via queries alone

* Record max BDE in spectra results table

* Calculate frequency in the absence of scores (for non-MSn method)

* Retain substructures does not cause substructures not to be initially recorded

* Add additional scoring metrics

* Update results db test data

* Define ppm error and valence of fragment prior to re-ordering

* Configure checks on recording of putative structure information

* Calculate scores at substructure combination level

* Convert True to 1 and False to 0 for conversion to SQLite boolean type

* Index results DB

* Use a loop in place of pool.map

* Minor performance improvements

* Merge minor performance improvements

* Use the minimum absolute error for getting possible fragment ions

* Add separate absolute error options for MSn peak and full structure

* Use 0.005 for abs_error_precursor

* Drop indexes before inserting into results DB

* Add results table index on ms_id_num and structure_smiles

* Update results DB tests

* Add table for generating unique structure smiles IDs

* Calculate cosine spectrum similarity

* Allow for the specification of weights for the results database scoring calculations

* Aggregate structure scores but force floating point division

* Select fragment and substructure id when calculating results scores for the correlated query

* Update results DB tests with updated scores

* Don't create indexes until structure scoring

* Don't include valence=0 substructures in the substructure database

* Add max BDE parameter for building

* Remove redundant connectivity graphs

* Update data to test filter records function

* Update dictionary pickle with Python 3.7

* Update file header

* Update contact information

* Update setup.py

* Update tests for RDKit changes

* Update README

* Keep functioning buttons

* Update testing workflow

* Use python 3.7

* Remove unused dependencies

* Use only the channel conda-forge

* Add pillow and pyqt dependencies

* Remove list definition in function arguments

* Add algorithms test

* Merge database tests into single file

* Restructure modules

* Restructure tests

* Update outdated imports

* Omit notebooks from coverage

Co-authored-by: Ralf Weber <[email protected]>