CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.0.6] - 2026-05-29

Fixed

• Recurrence sets with master in non-first position were rejected: RFC 5545 does not mandate that the master VEVENT (carrying RRULE) appears first in a VCALENDAR. Some CalDAV servers (notably calendar.mail.ru) emit override VEVENTs (with RECURRENCE-ID) before the master. The previous validation assumed components[0] was always the master and raised ValueError for any other ordering, which caused caldav._filter_search_results to silently drop all expanded recurring instances. The master is now located by content (has RRULE, no RECURRENCE-ID) regardless of position, and the list is reordered so it sits at index 0 for downstream expansion. Also reduced duplicated error-check code in the same function.

• check_component() mutated self on every call: The method resolved None fields (include_completed, todo, event, journal) and normalised start/end/alarm_start/alarm_end date objects to datetimes by writing back to self. This made a Searcher instance stateful: calling check_component() twice could produce different results, and reusing a Searcher across multiple search operations (e.g. in the python-caldav library) could silently change behaviour after the first call. All these values are now computed as local variables inside check_component() and threaded to the internal filter methods via new keyword parameters (_start, _end, _alarm_start, _alarm_end, _include_completed), keeping self immutable throughout. (Triggered by python-caldav/caldav#650)

• VTODO with only COMPLETED (no DTSTART/DUE/CREATED) crashed on time-range search: A time-range filter on a VTODO that has COMPLETED but no DTSTART, DUE, or CREATED would crash. The RFC 4791 §9.9 matrix was also re-read and the implementation now correctly handles all VTODO timestamp combinations: a task with only CREATED is treated as open-ended (matches any search interval ending after CREATED); a task with only COMPLETED is treated as a point event; and a task with both CREATED and COMPLETED uses them as the time bounds when no DTSTART/DUE is present.

[1.0.5] - 2026-02-19

Changes

Replace Poetry with Hatch. Version 1.0.4 never got as far as pypi due to a silly lock file problem.

[1.0.4] - 2026-02-19

Fixed

• undef filter for CATEGORIES was broken with icalendar >= 6.x: icalendar 6.x provides a default empty vCategory object for the CATEGORIES property even when it is not explicitly present in the iCalendar data, making "categories" in component always return True. This caused searches like no_category=True (or add_property_filter("categories", ..., operator="undef")) to incorrectly filter out all events instead of only those that actually have categories. Fixed by checking the computed category set (empty → not defined) rather than the unreliable in test.

• undef filter for DTEND was broken for recurring all-day events: recurring_ical_events adds a computed DTEND = DTSTART + 1 day to individual occurrences of all-day recurring events, even when the master event does not have an explicit DTEND. This caused searches for events without DTEND (e.g. no_dtend=True) to incorrectly filter out recurring all-day events. Fixed by skipping undef checks when filtering expanded recurrence occurrences whose master element has already passed the undef check.

[1.0.3] - 2025-11-30

Fixed

Support for RFC7986-style CATEGORIES split over multiple lines. #6

• The sort method did not handle lists very well. In particular, RFC 7986 section 5.6 defines that the categories field may be split over multiple lines in the icalendar object, this could cause an exception to be raised. Now lists are sorted and converted to a comma-separated string prior to sorting.
• The filter method would also run into problems, it has also been fixed

[1.0.0] - 2025-11-27

Added

• filter() method: Implemented filtering for lists of Calendar/Component objects

  • Accepts list[Calendar | Component] as input
  • Optional split_expanded parameter for splitting expanded recurrences into separate Calendar objects
  • Returns new list without modifying input (immutability guarantee)
  • Preserves Calendar properties and VTIMEZONE components
  • Applies all configured property filters and component type filters
  • Handles recurrence expansion when self.expand is True
  • Deep copies all returned objects for safety

• sort() method: Implemented sorting for lists of components/calendars

  • Accepts list[Component | CalendarObjectResource] as input
  • Returns new sorted list without modifying input
  • Uses configured sort keys (via add_sort_key())
  • Supports multiple sort keys with proper precedence
  • Supports case-sensitive/case-insensitive sorting
  • Supports reversed sorting
  • Handles Calendar, Component, and CalendarObjectResource objects

• filter_calendar() method: Implemented filtering of subcomponents within a Calendar object

  • Filters all subcomponents (events/todos/journals) within a single Calendar
  • Returns new Calendar with only matching subcomponents, or None if no matches
  • Preserves VTIMEZONE components
  • Preserves all Calendar-level properties
  • Handles recurrence expansion when self.expand is True
  • Deep copies to ensure immutability

• sort_calendar() method: Implemented sorting of subcomponents within a Calendar object

  • Sorts all subcomponents (events/todos/journals) within a single Calendar
  • Preserves VTIMEZONE components in their original position
  • Returns new Calendar without modifying input
  • Preserves all Calendar-level properties
  • Deep copies to ensure immutability

• Comprehensive test suites:

  • 15 tests for filter() method covering filtering, expansion, splitting, and edge cases
  • 5 tests for filter_calendar() method covering basic filtering, no matches, timezones, expansion, and mixed types
  • 12 tests for sort() and sort_calendar() methods covering various sorting scenarios

Changed

• Core functionality now complete: The library's main methods (filter(), filter_calendar(), sort(), sort_calendar()) are now fully implemented and no longer stubs
• README updated: Added usage examples for filter() and sort() methods, updated status section to reflect completion of core functionality

Notes

• This is the first stable 1.0 release
• The API is considered stable and will follow semantic versioning going forward
• Breaking changes in future versions will require a major version bump (2.0, 3.0, etc.)

[0.5.0] - 2025-11-26

Changed

• BREAKING: Collation system redesigned: Simplified collation architecture to use three collation strategies (SIMPLE, UNICODE, LOCALE), each supporting a case_sensitive parameter
  • Removed Collation.BINARY - use Collation.SIMPLE with case_sensitive=True instead
  • Removed Collation.CASE_INSENSITIVE - use Collation.SIMPLE with case_sensitive=False instead
  • All collation strategies (SIMPLE, UNICODE, LOCALE) now accept case_sensitive boolean parameter
  • UNICODE and LOCALE collations can now be case-sensitive or case-insensitive
  • PyICU collator strength is configured based on case_sensitive parameter (TERTIARY for case-sensitive, SECONDARY for case-insensitive)

Migration Guide

• Replace collation=Collation.BINARY with collation=Collation.SIMPLE, case_sensitive=True
• Replace collation=Collation.CASE_INSENSITIVE with collation=Collation.SIMPLE, case_sensitive=False
• UNICODE/LOCALE collations now require explicit case_sensitive parameter if you want case-sensitive behavior

[0.4.1] - 2025-11-26

Added

• UNICODE and LOCALE collation support for "==" operator: The == operator now properly supports UNICODE and LOCALE collations using ICU sort key comparison. Previously only BINARY and CASE_INSENSITIVE collations were supported for exact match operations.

Fixed

• PyICU-based tests now pass in CI environments where PyICU is installed

[0.3.1] - 2025-11-26

Some wheel files etc that shouldn't be in the repository has accidentally been added. Either I've done some silly mistake while sleeping, or perhaps it was a mistake giving the AI access to my computer.

[0.3.0] - 2025-11-26

Categories are a mess. I've tried to make a predictable behaviour by making a distinction between categories (handling the property filter value as a list that should match or be a subset of the event category set) and category (handling the property filter value as a string that should match towards at least one of the categories in the event).

Added

• "category" (singular) property filter: New virtual property for substring matching within category names
  • "category" with "contains" operator does substring matching (e.g., "out" matches "outdoor")
  • "category" with "==" operator matches if value exactly equals at least one category name
  • Commas in filter values are treated as literal characters, not split into multiple categories
  • Complements existing "categories" (plural) which does exact category name matching

Fixed

• Category filtering now properly respects collation settings for case-insensitive matching
• Category "contains" operator now correctly checks for subset (exact category names) rather than substring matching
• Category "==" operator now properly supports case-insensitive exact matching
• Text properties with "==" operator now support case-insensitive comparison when case_sensitive=False
• Category preprocessing now correctly handles tuples with comma-separated strings

[0.2.0] - 2025-11-24

Text filtering can be done both case-sensitive and case-insensitive - we need to support both of them. And we also may need to support various collations. DISCLAIMER: While I could easily add case insensitive/sensitive filtering myself, as soon as collations was tossed into the equation I considered the AI could do it better and faster than me. So 0.2.0 is AI-generated code, but curated by a human.

Added

• Collation support for text searches and sorting: Added comprehensive support for both case-sensitive and case-insensitive text comparisons
  • New case_sensitive parameter in add_property_filter() and add_sort_key() methods for simple API
  • New Collation enum for power users with support for BINARY, CASE_INSENSITIVE, UNICODE, and LOCALE collation strategies
  • Optional PyICU integration for advanced Unicode and locale-aware collation (install with pip install 'icalendar-searcher[collation]')
  • Collation support for all text properties (SUMMARY, LOCATION, DESCRIPTION, etc.)
  • Collation support for CATEGORIES property with special handling
  • Graceful fallback when PyICU is not installed

Changed

• BREAKING: Default text search behavior is now case-sensitive (was case-insensitive)
• Category searches are now case-sensitive by default (was implicitly case-sensitive, now explicit)

Fixed

• Improved type handling in property filters to correctly distinguish between text properties and other types

[0.1.10] - Previous Release