Skip to content

Latest commit

 

History

History
1121 lines (865 loc) · 36.6 KB

CHANGELOG.rst

File metadata and controls

1121 lines (865 loc) · 36.6 KB

Changelog

This project adheres to Semantic Versioning.

Unreleased

  • Any unreleased changes can be viewed in the latest version documentation changelog.

0.39.0 (2024-11-15)

Removed

  • Removed all support for base-2 formatting. The sciform package will focus on base-10 scientific formatting, more relevent for the physical sciences, unless there is demand for base-2 or base-n formatting. We encourage users to consider the prefixed package if they have requirements for base-2 formatting. [#201]

0.38.2 (2024-11-02)

Fixed

  • Corrected a typo in a previous release.

0.38.1 (2024-11-02)

Added

  • Documented examples demonstrating the interplay between sciform formatting the decimal module configuration. [#183]
  • Added an example demonstrating how users can implement a "fallback" formatter for cases when users want different formatting behavior applied when the uncertainty is invalid or not present. [#177]

0.38.0 (2024-11-02)

Added

  • Digits-past-the-decimal formatting (round_mode="dec_place") is now supported for value/uncertainty pairs. [#170]
  • round_mode="all" and round_mode="pdg" options can now be accessed in the FSML using "A" and "P" flags. [#192]
  • Added a dark theme option (default on) for readthedocs documentation.
  • Added code coverage breakdown for unit and feature tests to codecov CI.

Changed

  • [BREAKING] Rounding configuration has been refactored. Previously round_mode accepted only "sig_fig" and "dec_place" string literals, all digits rounding mode was selected by settings ndigits = AutoDigits and PDG rounding mode was selected by setting pdg_sig_figs = True. Now round_mode accepts "sig_fig", "dec_place", "pdg" and "all" string literals, ndigits only accepts integers, and the pdg_sig_figs option is removed. round_mode defaults to "all" and ndigits defaults to 2. [#185]
  • [BREAKING] Previously exp_val and ndigits accepted the enums AutoExpVal and AutoDigits. Now exp_val accepts the string literal "auto" and ndigits only accepts integers. [#178, #185]
  • Previously, significant figure rounding with a value of 0 would result in trailing zeros being added in the formatted output. For example, format(SciNum(0), "!3f") would give "0.00" or format(SciNum(0, float("nan")), "!2f") would give "0.0 ± nan". This is not reasonable because zero doesn't have any significant figures, so it doesn't make sense to add more trailing zeros to indicate additional "fake" significant figures. Now if zero is used for significant figure rounding it always appears directly as "0". E.g. format(SciNum(0), "!3f") gives "0" and format(SciNum(0, float("nan")), "!2f") gives "0 ± nan". Note that, as before, trailing zeros may still be added to a zero value if the uncertainty is less than one, e.g. format(SciNum(0, 0.0012), "!2f") gives "0.0000 ± 0.0012". [#189]
  • Previously the backend FinalizedOptions class ran a validation check on itself after initialization. This check has been removed in favor of not complicating the validation code to handle the above change to exp_val and ndigits. Users never directly instantiate FinalizedOptions so this will hopefully be no problem. InputOptions continue to be validated because this is the direct result of user input. PopulatedOptions also must continue to be validated because options conflicts due to the combination of user input with global options can arise at options population time.
  • Update ruff version in pre-commit config.

REMOVED

  • The pdg_sig_figs options has been removed. This option is now configured by setting round_mode="pdg".

FIXED

  • The ruff version in the github CI now matches the ruff in the local precommit configuration.

0.37.0 (2024-03-07)

Added

  • Added support for Python 3.8.

Changed

  • Move ruff configuration out of pyproject.toml into ruff.toml. [#163]

0.36.0 (2024-03-07)

Added

  • Added many unit test to supplement the feature existing feature tests. [#102]
  • Added much more input validation and corresponding messaging.
    • Checks on input types and values.
    • Extra translations dictionaries are now checked so that keys must be integers and values must consist of only English alphabetic characters. [#157]

Changed

  • Major code reorganization. [#152]
    • Move modules containing public interfaces into an api sub-package.
    • Break the bulky format_utils module into multiple modules, now in a format_utils sub-package.
    • Collect the main formatting algorithms into a formatting sub-package.
    • Sort tests into feature and unit tests.
  • Some utility code refactoring.

Fixed

  • Fixed a bug where uncertainties between the Particle Data Group uncertainties thresholds would erroneously result in ValueError being raised when formatted with pdg_sig_figs=True. [#164]

0.35.0 (2024-02-16)

Added

  • The Formatter formatting method and the SciNum constructor now accept formatted inputs for the value and optional uncertainty inputs. E.g. formatter("123.456(7)e-03") or SciNum("24 +/- 2") are now valid inputs. [#104]

  • Added the paren_uncertainties_trim option. The previous behavior was paren_uncertainties_trim=True. Now paren_uncertainties_trim=False allows a more verbose presentation of the uncertainty in paren_uncertainty mode in which leading zeros and separator characters are not stripped from the string. E.g. paren_uncertainties_trim=True will give

    123.002 3(21)

    while paren_uncertainties_trim=False will give

    123.002 3(0.002 1)

  • Added value and uncertainty attributes to the FormattedNumber class.

  • Added badge for Zenodo.

Removed

  • [BREAKING] Removed the paren_uncertainties_separators option. This option made it possible (when False) to optionally strip all separator characters, including the decimal separator, from the uncertainty in paren_uncertainty mode. This lead to the possibility of value/uncertainty pairs like

    123 456.789 8 ± 123.456 7

    being represented as

    123 456.789 8(1234567)

    sciform will now display this as

    123 456.789 8(123.456 7)

    if paren_uncertainty_strip=False or

    123 456.789 8(123.4567)

    if paren_uncertainty_strip=True, but always retaining the decimal separator. In most cases many fewer significant digits of the uncertainty are displayed and the resulting outputs don't look as egregious when the decimal separator is stripped. Nonetheless, given that more outputs look better when the decimal is retained and that there is no official BIPM guidance on how parentheses should handle cases when the uncertainty digits span decimal or other separator characters, sciform will not presently provide an option to strip the decimal separator character.

Changed

  • Previously, when using paren_uncertainty=True, redundant parentheses were included around the value and uncertainty numbers if there was an ASCII exponent (e.g. e+02) or in percent formatting mode. E.g. outputs would look like (32.4(1.2))e+02 or (32.4(1.2))%. Now these redundant parentheses are excluded so outputs look like 32.4(1.2)e+02 or 32.4(1.2)%. This is consistent with how the uncertainties package handles these cases. The extra parentheses were originally included for increased clarity, but the extra parentheses only clutter the output and there is sufficient clarity without them. This change eliminates an issue where the redundant parentheses were erroneously included or excluded after LaTeX/HTML/ASCII output conversion. [#145]

Fixed

  • Previously, when formatting individual Decimal input values, the values were always normalized at an early stage in formatting. This meant that even if ndigits=AutoDigits then Decimal("1.0") would be formatted the same as Decimal("1.00"). However, for value/uncertainty formatting, Decimal input to the uncertainty was not necessarily normalized at an early stage. This meant that with ndigits=AutoDigits, an uncertainty of Decimal("1.0") would be formatted to the tenths decimal place while an uncertainty of Decimal("1.00") would be formatted to the hundredths place. This behavior was inconsistent and undocumented. Now all Decimal inputs are immediately normalized before any formatting. [#148]
  • Fixed the behavior around the sign symbols for zero and non-finite inputs. Previously 0 was treated as positive for the sake of resolving its sign symbol, the sign of infinite numbers was preserved but +inf did not respect the "+" and " " sign modes, and nan never had a sign but also never had an extra character added for "+" or " " sign modes. Now both 0 and nan are treated as having no sign. In both "+" and " " sign modes 0 and nan are prepended by a space. The sign of infinite numbers is retained as before, but now formatting of these numbers respects the sign mode. [#147]

0.34.1 (2024-02-10)

Added

  • Updated the readme to reflect completion of the PyOpenSci review.

0.34.0 (2024-02-04)

Added

  • The Formatter now exposes the input_options and populated_options attributes. The input_options attribute holds an InputOptions object which stores a record of the input options passed into the Formatter. The populated_options attribute returns a PopulatedOptions object which shows the complete set of populated options which will be used for formatting after merging with the global options. Note that the populated_options attribute is re-calculated each time it is access so that it reflects the current global options. Both the InputOptions and PopulatedOptions objects can be used to provide string representations of the options, or provide programmatic access to the options via either attribute access or the as_dict() methods. [#110]
  • The PopulatedOptions used during formatting of a given FormattedNumber instance are stored on that instance for future reference.
  • Added get_default_global_options().
  • Now integer 0 can be passed into left_pad_char to get the same behavior as string "0".
  • Added tests for docstrings.

Changed

  • [BREAKING] Renamed functions for configuring global options:
    • set_global_defaults() -> set_global_options()
    • reset_global_defaults() -> reset_global_options()
    • GlobalDefaultsContext() -> GlobalOptionsContext()
  • Refactored backend options handling code. Previously, UserOptions were rendered into RenderedOptions. During rendering the global options were appropriately merged in and some string literal options were replaced with enums for internal use. These two classes were private. Now there are InputOptions (which try to faithfully record user input), PopulatedOptions (which capture the result of merging the global options into the input options, but still using user-friendly string representations of all options), and FinalizedOptions (which use the internal enum representations of certain options). The InputOptions and PopulatedOptions are now public while the FinalizedOptions is still private to shield the enum representations from the users. This sizable refactor was precipitated by the publicizing of the options. [#110]

Removed

  • [BREAKING] Removed print_global_defaults() in favor of get_global_defaults() which now returns a PopulatedOptions object which can be printed by the user if desired.

Fixed

  • Fixed a bug where SciNum formatting resulted in str objects instead of FormattedNumber objects.

0.33.0 (2024-01-31)

Added

  • Added the FormattedNumber class. This class is a subclass of str and is now returned by the Formatter instead of str. The FormattedNumber class allows post-conversion to ASCII, HTML, and LaTeX formats. [#114]
  • Added separate flags for code coverage reports for each python version.

Changed

  • In addition to removing the latex option from the Formatter in favor of the introduction of the FormattedNumber class, the LaTeX conversion algorithm has been slightly modified.

    • Left and right parentheses are no longer converted to "\left(" and "\right)" due to introducing strange spacing issues. See Spacing around \left and \right.
    • Previously spaces within the sciform output were handled inconsistently and occasionally required confusing extra handling. Now any spaces in the input string are directly and explicitly converted into math mode medium spaces: "\:".
    • "μ" is now included in the math mode \text{} environment and converted to "\textmu".

  • [BREAKING] Renamed fill_char to left_pad_char. [#126]

  • Slimmed down [dev] optional dependencies and created [examples] optional dependencies. The former includes development tools, while the latter includes the heavy-weight requirements needed to run all the examples, including, e.g. jupyter, scipy, etc.

  • Cleaned up and improved github actions for testing and linting/formatting. [#136]

    • Use unittest and coverage instead of pytest.
    • The requirements to run the automation match the [dev] optional dependencies.
    • Cache pip requirements to avoid unnecessarily downloading dependencies.
    • Remove a defunct blackdoc test. Hopefully this can be replaced when ruff provides functionality for formatting .rst files.

Fixed

  • Fixed a bug where value/uncertainty pairs formatted in the "parts_per" format with zero exponent would appear with redundant parentheses, e.g. "(1.2 ± 0.1)". [#130]

Removed

  • [BREAKING] Removed the latex option in favor of the introduction of the FormattedNumber.as_latex() method. This removal simplifies the formatting algorithm by separating LaTeX formatting from other tasks like exponent string resolution. The latex option also introduced a potential confusion with the superscript option, which had no effect when latex=True.

0.32.3 (2024-01-11)

Added

  • Added more PyPi classifiers.

0.32.2 (2024-01-11)

Added

  • Expanded the "Under Construction" section of the readme and the "How to Contribute" section of the project page. Changes included adding links to the sciform feedback survey.
  • Added examples in the documentation demonstrating how sciform formatting can be mapped over collections of numbers. [#120]

Changed

  • Refactor backend mode literal (used for typing) and enum (used internally for tracking options) object names so that e.g. SignMode -> SignModeEnum and UserSignMode -> SignMode. [#111]

0.32.0 (2024-01-10)

Added

  • Previously it was impossible to configure pdg_sig_figs=True together with ndigits!=AutoDigits. This combinations resulted in an exception. Now behavior has been defined and implemented for this combination. For single value formatting the value of pdg_sig_figs is always ignored. For value/uncertainty formatting ndigits is ignored if pdg_sig_figs=True. The behavior for pdg_sig_figs=False is unchanged. [#73]

Removed

  • [BREAKING] Removed global_add_c_prefix, global_add_small_si_prefixes, global_add_ppth_form, global_reset_si_prefixes, global_reset_iec_prefixes, and global_reset_parts_per_forms. These options are redundant with set_global_defaults and GlobalDefaultsContext and make the extra translations dictionaries more confusing to understand. [#97]

Changed

  • [BREAKING] Previously 12.3 would format as "12.3e+00" when using parts per formatting mode. Now, when using parts per formatting mode, the e+00 exponent is translated to be an empty string so that 12.3 would format as "12.3". [#99]

0.31.1 (2024-01-06)

Removed

  • [BREAKING] Removed the SciNumUnc class. Now the SciNum class can be used with an optional second positional argument to specify the uncertainty associated with a number.
  • [BREAKING] Remove separator configuration from the FSML. These options made the FSML to cumbersome and led to confusing (if not incorrect) conflicts with the round mode symbol. Now all separator configuration needs to be done by setting the global format options or using the global format options context manager. [#29]

Added

  • Added annotated examples demonstrating the FSML.
  • Added more documentation for contributing developers.
  • Added pre-commit configuration.

Changed

  • [BREAKING] Renamed multiple options.

    • top_dig_place renamed to left_pad_dec_place.
    • superscript_exp renamed to superscript.
    • bracket_unc renamed to paren_uncertainty.
    • bracket_unc_remove_seps renamed to paren_uncertainty_separators. This change is associated with a a reversal of the Boolean logic on the option.
    • val_unc_match_widths renamed to left_pad_matching.
    • unc_pm_whitespace renamed to pm_whitespace.

  • [BREAKING] Previously specifying any left pad decimal place using the sciform FSML resulted in setting left_pad_matching=True so that print(f"{SciNum(123.456, 0.789):0}") resulted in "123.456 ± 000.789". Now the FSML has no impact on left_pad_matching. Now, similar to many other options, the global setting for left_pad_matching will always be used when formatting using the FSML. Under the default global options (left_pad_matching=False) print(f"{SciNum(123.456, 0.789):0}") results in "123.456 ± 0.789".

  • Implemented ruff linting and formatting in codebase and integration automation.

  • Refactored code for adding separators.

  • Refactored formatting and formatting utilities to simplify functions and make the algorithm easier to follow.

  • More aggressively filter JetBrains .idea/ folder from version control.

Fixed

  • Fixed a bug involving removing separators in parentheses uncertainty mode when at least one of the value and uncertainty were non-finite.

0.30.1 (2023-11-24)

Fixed

  • Fixed Changelog.

0.30.0 (2023-11-24)

Changed

  • [BREAKING] Remove the FormatOptions class from the user interface. Now users configure Formatter instances by passing the formatting keyword arguments into the Formatter constructor directly. Global configuration via set_global_defaults() or the GlobalDefaultsContext is also done by passing formatting keywords directly. This change reduces the amount of boilerplate code and keystrokes needed to use sciform.
  • [BREAKING] Options such as exp_mode and exp_format were previously configured using Enum objects such as ExpMode or ExpFormat. Now these options are configured using string literals. This change also reduces the amount of boilerplate code and keystrokes needed to use sciform.
  • Clean up print_global_defaults output. This is the start of an effort to improve interface for getting and printing current format options.

Added

  • Added code of conduct.
  • Added contributing guidelines.
  • Added Python 3.12 to automated testing.

Fixed

  • Cleaned up API documentation.
  • Fixed a bug where the repr for FormatOptions would return a string containing information about the global format options rather than about the specific FormatOptions instance. [#75]
  • Fixed an issue that was causing Github actions code coverage report to not actually check code coverage. [#84]

Removed

  • [BREAKING] Removed the unicode_pm feature which allowed toggling between using '+/-' or '±' in value/uncertainty strings. Previously unicode_pm defaulted to False so that '+/-' was the default behavior. Now the default behavior is to use '±' and there is no way to change to the old '+/-' behavior. [#10]

0.29.1 (2023-10-22)

Fixed

  • Fixed a bug where bracket uncertainties erroneously appeared as empty parentheses for zero or non-finite uncertainties. [#66]
  • Fixed a bug where the exponent value was erroneously calculated from the uncertainty rather than the value when the value was negative (but larger in magnitude than the uncertainty). [#68]
  • Fixed a bug where certain leading digits were erroneously not stripped from the uncertainty when using bracket uncertainty with negative values. [#68]
  • Fixed a bug where the value was erroneously being rounded according to the PDG rounding rules when pdg_sig_figs=True, the uncertainty was zero or non-finite, and the value was positive. [#71]
  • Fixed a bug where a spurious error was raised when pdg_sig_figs=True, the uncertainty was zero or non-finite, and the value was zero or negative. [#65]

Changed

  • Replace -e . with . in requirements.txt. There is no need to install sciform in editable mode for code automation routines.

0.29.0 (2023-09-05)

Changed

  • Previously, when using bracket_unc=True with any exponent string (such as e-06, μ or ppm), the value and uncertainty were always wrapped in parentheses, e.g. (1.03(25))e-06, (1.03(25)) μ or (1.03(25)) ppm. Now, when using bracket_unc=True with prefix or parts-per exponent format modes, if the exponent is replaced with an alphabetic replacement, then the value and uncertainty are no longer wrapped in parentheses, e.g. 1.03(25) μ and 1.03(25) ppm. This is consistent with BIPM Guide Section 7.2.2. Specifically, any time bracket_unc=False the value and uncertainty are always wrapped in parentheses, and any time notation like e+02 or b+02 is used to indicate the exponent then the value and uncertainty are always wrapped in parentheses.

Fixed

  • Correct fit_plot_with_sciform.py example script to use new exp_format=ExpFormat.PREFIX instead of old prefix_exp=True.

Improved

  • Documentation improvements including typos and neatening up changelog.

0.28.2 (2023-08-31)

Improved

  • General wording and grammar improvements throughout documentation.
  • Include more usage examples in the examples documentation in addition to referring the reader to the test suite.

Fixed

  • Fixed a bug when using pdg_sig_figs with uncertainties larger than about 1000 by cleaning up Decimal math.
  • Previously, when formatting using the format specification mini-language, if the prefix exponent format flag was omitted then the exponent format was forced to ExpFormat.STANDARD rather than None. This meant that it was impossible, using the format specification mini-language combined with global configuration options, to set ExpFormat.PARTS_PER. Now when the prefix flag is omitted exp_format is set to None so that it will be populated by the global default option. In the future a flag may be added to select "parts-per" formatting using the format specification mini-language.

0.28.1 (2023-08-28)

  • Make FormatOptions inputs Optional so that None inputs pass type checks.
  • Write format-specification mini-language documentation to refer to existing format options documentation to avoid documentation duplication.
  • Setup test coverage analysis automation and upload report to codecov.
  • Add package status badges to readme.
  • Test against Python 3.11.
  • List supported Python versions in pyproject.toml classifiers.

0.28.0 (2023-08-27)

  • [BREAKING] Replace prefix_exp and parts_per_exp options with an exp_format option which can be configured to ExpFormat.STANDARD, ExpFormat.PREFIX or ExpFormat.PARTS_PER.
  • Previously formatting a non-finite number in percent mode would always display a '%' symbol, e.g. '(nan)%'. Now the brackets and '%' symbol will be omitted unless nan_inf_exp=True.
  • In latex=True mode there is now a space between the number and a prefix or parts-per translated exponent. For value/uncertainty formatting the space is still absent. For latex=False there is still always a space for number and value/uncertainty formatting before the translated exponent string.
  • In latex=True mode 'nan' and 'inf' strings are now wrapped in '\text{}'.
  • Refactored code for resolving exponent strings.
  • Added more unit tests to reach 100% test coverage. Mostly added test cases for invalid internal inputs.
  • Raise NotImplementedError when attempting value/uncertainty formatting with binary exponent modes. Rounding and truncating are not properly implemented in binary mode yet.

0.27.4 (2023-08-25)

  • Setup github action to automatically build and publish on release.

0.27.3 (2023-08-23)

  • Added Unreleased section to changelog.
  • Removed version from source code. Project version is now derived from a git version tag using setuptools_scm.
  • Stopped encouraging import FormatOptions as Fo.

0.27.2 (2023-08-20)

  • Add __repr__() for FormatOptions and RenderedFormatOptions.

0.27.1 (2023-08-18)

  • Add examples/ folder to hold example scripts used in the documentation as well as the input data for these scripts and their outputs which appear in the documentation.
  • Remove extra readthedocs.yaml file.

0.27.0 (2023-08-18)

  • [BREAKING] Rename AutoRound to AutoDigits. This is because, e.g., ndigits=AutoDigits sounds more correct than ndigits=AutoRound. Furthermore, AutoRound could likely be confused as being an option for round_mode, which it is not.

0.26.2 (2023-08-18)

  • Fix a bug where illegal options combinations could be realized at format time when certain global default objects were merged into certain user specified options. The bug is fixed by re-checking the options combinations after merging in the global defaults but before formatting.

0.26.1 (2023-08-18)

  • Add unit tests, increase test coverage.

0.26.0 (2023-08-15)

  • [BREAKING] Rename some format options to make their usage more clear.

    • exp to exp_val
    • precision to ndigits
    • RoundMode.PREC to RoundMode.DEC_PLACE
    • AutoExp to AutoExpVal
    • AutoPrec to AutoRound

  • Raise more exceptions for incorrect options combinations.

    • Raise an exception when using pdg_sig_figs with a user-supplied exp_val.
    • Raise exceptions instead of warnings for invalid user-supplied exp_val in get_mantissa_base_exp().

  • Minor refactor to GlobalDefaultsContext.

  • Documentation:

    • Update documentation to reflect name changes above.
    • Better centralization of float/Decimal information.
    • Better explanations of AutoExpVal and AutoRound behavior.
    • More accurate descriptions of some invalid options combinations.

0.25.2 (2023-08-11)

  • Update roadmap

0.25.1 (2023-08-10)

  • Refactor get_pdg_round_digit() into a dedicated function.

0.25.0 (2023-08-02)

  • [BREAKING] template option removed from FormatOptions constructor. New FormatOptions instances can be constructed from two existing FormatOptions instances using the merge() method.
  • Minor documentation improvements.

0.24.0 (2023-07-30)

  • [BREAKING] percent mode is now accessed via an exponent mode, ExpMode.PERCENT. There is no longer a percent keyword argument.

0.23.0 (2023-07-29)

  • [BREAKING] Users now construct FormatOptions objects which they pass into Formatter objects and global configuration functions. Formatter and global configuration functions no longer accept bare keyword arguments to indicate formatting options.
  • [BREAKING] Formatter now resolves un-filled format options from the global defaults at format time instead of initialization time. This is consistent with the previous behavior for SciNum and SciNumUnc objects.
  • Change pyproject.toml description

0.22.2 (2023-07-27)

  • Add .readthedocs.yaml and update documentation requirements.txt for reproducible documentation builds.

0.22.1 (2023-07-27)

  • Fix a date typo in the changelog for the entry for version 0.22.0.

0.22.0 (2023-07-27)

  • [BREAKING] Rename sfloat to SciNum and vufloat to SciNumUnc

  • [BREAKING] SciNum instances do not support arithmetic operations the same way sfloat instances did. This functionality was removed for two reasons. First, SciNum uses Decimal to store its value instead of float and configuring SciNum to behave as a subclass of Decimal would require added complexity. Second, A decision has been made to keep the sciform module focussed solely on formatting individual numbers or pairs of numbers for early releases. Convenience functionality outside of this narrow scope will be considered at a later time.

  • Favor Decimal methods over float methods in internal formatting algorithm code.

  • Documentation

    • Remove float-based language fom documentation.
    • Include a discussion in the documentation about Decimal versus float considerations that may be important for users.
    • Various minor revisions and edits. Notably a typo in the version 0.21.0 changelog entry that reversed the meaning of a sentence was corrected.
    • Add "under construction" message to README.

0.21.0 (2023-07-22)

  • Use Decimal under the hood for numerical formatting instead of float. Decimal instances support higher precision than float and more reliable rounding behavior.
  • Update particle data group uncertainty rounding unit tests since edge cases are now handled property as a result of adopting Decimal.
  • Minor cleanup of sfloat arithmetic functions.

0.20.1 (2023-06-24)

  • Refactor unit tests to use lists and tuples instead of dicts. Literal dicts allow the possibility for defining the same key (test case) with different values, only the latest of which will actually be tested. The refactoring ensures all elements of the test lists will be tested.
  • Refactor sfloat and vufloat __format__() functions to call format_float() and format_val_unc() directly instead of creating a Formatter object first.

0.20.0 (2023-06-22)

  • Support passing None as a value into extra_si_prefixes, extra_iec_prefixes, or extra_parts_per_forms to prevent translation of a certain exponent value. This may be useful for suppressing ppb or similar local-dependent "parts per" translations.
  • [BREAKING] Change the bracket uncertainty flag in the FSML from 'S' to '()'.
  • When an exponent translation mode is used in combination with Latex mode, the translated exponent will now be wrapped in a Latex text mode: e.g. \text{Mi}.
  • Link to test cases on examples page.

0.19.0 (2023-06-22)

  • Add python-package.yaml github workflows. Allows automated testing, doc testing, and flake8 scans during github pull requests.
  • Minor flake8 cleanup

0.18.1 (2023-06-21)

  • Documentation improvements

0.18.0 (2023-06-19)

  • Add Particle Data Group significant figure auto selection feature, documentation, and tests.
  • [BREAKING] Use the larger of value or uncertainty to resolve the exponent when formatting value/uncertainty pairs. The previous behavior was to always use the value to resolve the exponent, but this behavior was not convenient for the important use case of zero value with non-zero uncertainty.
  • Expose AutoPrec and AutoExp sentinel classes so that users can explicitly indicate automatic precision and exponent selection.

0.17.1 (2023-06-19)

  • Code restructure to make formatting algorithm easier to follow including more verbose clarifying comments.
  • Minor documentation cleanup

0.17.0 (2023-06-19)

  • Add parts-per notation feature, documentation, and tests.
  • [BREAKING] Rename use_prefix option to prefix_exp.
  • Fix typos in binary IEC prefixes table.
  • Fix some cross links in documentation.

0.16.0 (2023-06-18)

  • Add latex option with documentation and tests.
  • Refactor exponent string conversion.
  • Percent mode for non-finite numbers.

0.15.2 (2023-06-18)

  • Fix a bug involving space filling and separators.

0.15.1 (2023-06-17)

  • Changelog formatting typo.

0.15.0 (2023-06-17)

  • Add superscript_exp option with documentation and tests.
  • Forbid percent mode unless using fixed point exponent mode.
  • Add PyPi link to readme.

0.14.0 (2023-06-17)

  • Add Changelog.
  • Add unicode_pm option with documentation and tests.