* enable python 3.14
* try on mac
* remove beta 4 disclaimer
* adds sleep before starting windows e2e tests
---------
Co-authored-by: Marcin Rudolf <rudolfix@rudolfix.org>
* enable 3.14 with orjson branch
* make example plugin a uv project
* post rebase pyproject update
* fix one dependency
update readme
* update readme about python 3.14
* add python 3.12 linting
* update locked versions to make project installable on py 3.12
* update flake8
* downgrade poetry for all tests relying on python3.8
* drop python 3.8
* enable python3.13
* copy test updates from python3.13 branch
* update locked sentry version
* pin poetry to 1.8.5
* install ibis outside of poetry
* rename to workflows for consistency
* switch to published alpha version of dlt-pendulum for python 3.13
* fix images
* add note to readme
* Add instruction on installing dlt via pixi and conda
* Update getting-started doc
* Add mention to use relevant package managers to install extra dependencies
* Fix typo
* Fix typo
* Update streamlit app screenshot
* Move installation instructions to installation.md
* Add a separate section for pixi and conda
* format examples
* add core functionality for scd2 merge strategy
* make scd2 validity column names configurable
* make alias descriptive
* add validity column name conflict checking
* extend write disposition with dictionary configuration option
* add default delete-insert merge strategy
* update write_disposition type hints
* extend tested destinations
* 2nd time setup (#1202)
* remove obsolete deepcopy
* add scd2 docs
* add write_disposition existence condition
* add nullability hints to validity columns
* cache functions to limit schema lookups
* add row_hash_column_name config option
* default to default merge strategy
* replace hardcoded column name with variable to fix test
* fix doc snippets
* compares records without order and with caps timestamps precision in scd2 tests
* defines create load id, stores package state typed, allows package state to be passed on, uses load_id as created_at if possible
* creates new package to normalize from extracted package so state is carried on
* bans direct pendulum import
* uses timestamps with properly reduced precision in scd2
* selects newest state by load_id, not created_at. this will not affect execution as long as packages are processed in order
* adds formating datetime literal to escape
* renames x-row-hash to x-row-version
* corrects json and pendulum imports
* uses unique column in scd2 sql generation
* renames arrow items literal
* adds limitations to docs
* passes only complete columns to arrow normalize
* renames mode to disposition
* saves parquet with timestamp precision corresponding to the destination and updates schema in the normalizer
* adds transform that computes hashes of tables
* tests arrow/pandas + scd2
* allows scd2 columns to be added to arrow items
* various renames
* uses generic caps when writing parquet if no destination context
* disables coercing timestamps in parquet arrow writer
---------
Co-authored-by: Jorrit Sandbrink <sandbj01@heiway.net>
Co-authored-by: adrianbr <adrian.brudaru@gmail.com>
Co-authored-by: rudolfix <rudolfix@rudolfix.org>
* Cancel previous CI runs for PRs when new commits pushed
* Small readme example pipeline code formatting
* Include matrix.os in group
* Use max-parallel: 3 for lint and common tests
* Skip matrix.os for linting
* Skip matrix.os for destination workflows
* Revert common tests workflow
* Parallelize linting
* Use only ubuntu to run linting
* Run qdrant tests only on linux
* Remove redundant qdrant workflow checks for windows os
* fix link
* bumps prerelease version 0.3.1a0
* fix link
* fixes links and moves some content to reference
* fixes links in the core library
---------
Co-authored-by: Marcin Rudolf <rudolfix@rudolfix.org>
