dataiter 1.4__tar.gz → 2.1__tar.gz

{dataiter-1.4 → dataiter-2.1}/.gitignore

@@ -2,8 +2,8 @@
 .env
 .envrc
 .pytest_cache
+.venv
 __pycache__
-benchmark-head.py
 benchmark-versions.csv
 benchmark-versions.ods
 build
@@ -11,6 +11,7 @@ dist
 doc/_build
 doc/comparison/comparison.html
 test.py
+tmp
 tmp.csv
+uv.lock
 validation/*.csv
-venv

{dataiter-1.4 → dataiter-2.1}/PKG-INFO

@@ -1,17 +1,19 @@
 Metadata-Version: 2.5
 Name: dataiter
-Version: 1.4
+Version: 2.1
 Summary: Simple, light-weight data frames for Python
 Project-URL: Homepage, https://github.com/otsaloma/dataiter
+Project-URL: Documentation, https://dataiter.readthedocs.io/
+Project-URL: Changelog, https://github.com/otsaloma/dataiter/blob/master/NEWS.md
 Author-email: Osmo Salomaa <otsaloma@iki.fi>
 License-Expression: MIT
 License-File: AUTHORS.md
 License-File: COPYING
-Requires-Python: >=3.11.0
+Requires-Python: >=3.12.0
 Requires-Dist: attd>=0.3
-Requires-Dist: numpy<3.0,>=2.4.0
-Requires-Dist: pyarrow>=16.0
-Requires-Dist: wcwidth>=0.1
+Requires-Dist: numpy<3.0,>=2.5.0
+Requires-Dist: pyarrow>=26.0
+Requires-Dist: wcwidth>=0.3.0
 Description-Content-Type: text/markdown
 
 # Simple, Light-Weight Data Frames for Python
@@ -24,8 +26,9 @@ Dataiter's **`DataFrame`** is a class for tabular data similar to R's
 is under the hood a dictionary of NumPy arrays and thus capable of fast
 vectorized operations. You can consider it to be a light-weight
 alternative to Pandas with a simple and consistent API. Performance-wise
-Dataiter relies on NumPy and Numba and is likely to be at best
-comparable to Pandas.
+Dataiter relies on NumPy and PyArrow and is close to the level of
+Pandas. Currently the notable slow part is string data and operations
+(Dataiter's NumPy StringDType being slower than Pandas' Arrow strings).
 
 ## Installation
 
@@ -35,18 +38,8 @@ pip install -U dataiter
 
 # Latest development version
 pip install -U git+https://github.com/otsaloma/dataiter
-
-# Numba (optional)
-pip install -U numba
 ```
 
-Recommended NumPy version is currently >= 2.4.0 due to various
-StringDType fixes that have landed in NumPy 2.2.1 and 2.4.0.
-
-Dataiter optionally uses **Numba** to speed up certain operations. If
-you have Numba installed, Dataiter will use it automatically. It's
-currently not a hard dependency, so you need to install it separately.
-
 ## Quick Start
 
 ```python
@@ -80,20 +73,23 @@ the differences and similarities in common operations.
 
 https://dataiter.readthedocs.io/en/stable/comparison.html
 
-For coding agents and LLMs, each documentation page is also available as
-Markdown by replacing `.html` with `.md` in the URL. There's also an
-`llms.txt` index and complete documentation as `llms-full.txt`. See
+### For Coding Agents
+
+Dataiter's API is small enough that the complete documentation fits in
+about 50k tokens. Coding agents can read all of it in one go and know
+the whole API before writing any code. This differs from bigger
+packages, like Pandas and Polars, which must fight against outdated
+training data and introduce their huge API piecemeal, hoping the agent
+knows what to ask for. For best results, have your agent download
+`llms-full.txt` and read it in full, rather than fetch it with a tool
+that summarizes web pages. Each documentation page is also available as
+Markdown by replacing `.html` with `.md` in the URL.
 
-- https://dataiter.readthedocs.io/en/stable/data-frame.md
-- https://dataiter.readthedocs.io/en/stable/llms.txt
 - https://dataiter.readthedocs.io/en/stable/llms-full.txt
+- https://dataiter.readthedocs.io/en/stable/data-frame.md
 
 ## Development
 
-To install a virtualenv for development, use
-
-    make venv
-
-or, for a specific Python version
+Install uv from https://docs.astral.sh/uv/ and then run
 
-    make PYTHON=python3.X venv
+    uv sync

{dataiter-1.4 → dataiter-2.1}/README.md

@@ -8,8 +8,9 @@ Dataiter's **`DataFrame`** is a class for tabular data similar to R's
 is under the hood a dictionary of NumPy arrays and thus capable of fast
 vectorized operations. You can consider it to be a light-weight
 alternative to Pandas with a simple and consistent API. Performance-wise
-Dataiter relies on NumPy and Numba and is likely to be at best
-comparable to Pandas.
+Dataiter relies on NumPy and PyArrow and is close to the level of
+Pandas. Currently the notable slow part is string data and operations
+(Dataiter's NumPy StringDType being slower than Pandas' Arrow strings).
 
 ## Installation
 
@@ -19,18 +20,8 @@ pip install -U dataiter
 
 # Latest development version
 pip install -U git+https://github.com/otsaloma/dataiter
-
-# Numba (optional)
-pip install -U numba
 ```
 
-Recommended NumPy version is currently >= 2.4.0 due to various
-StringDType fixes that have landed in NumPy 2.2.1 and 2.4.0.
-
-Dataiter optionally uses **Numba** to speed up certain operations. If
-you have Numba installed, Dataiter will use it automatically. It's
-currently not a hard dependency, so you need to install it separately.
-
 ## Quick Start
 
 ```python
@@ -64,20 +55,23 @@ the differences and similarities in common operations.
 
 https://dataiter.readthedocs.io/en/stable/comparison.html
 
-For coding agents and LLMs, each documentation page is also available as
-Markdown by replacing `.html` with `.md` in the URL. There's also an
-`llms.txt` index and complete documentation as `llms-full.txt`. See
+### For Coding Agents
+
+Dataiter's API is small enough that the complete documentation fits in
+about 50k tokens. Coding agents can read all of it in one go and know
+the whole API before writing any code. This differs from bigger
+packages, like Pandas and Polars, which must fight against outdated
+training data and introduce their huge API piecemeal, hoping the agent
+knows what to ask for. For best results, have your agent download
+`llms-full.txt` and read it in full, rather than fetch it with a tool
+that summarizes web pages. Each documentation page is also available as
+Markdown by replacing `.html` with `.md` in the URL.
 
-- https://dataiter.readthedocs.io/en/stable/data-frame.md
-- https://dataiter.readthedocs.io/en/stable/llms.txt
 - https://dataiter.readthedocs.io/en/stable/llms-full.txt
+- https://dataiter.readthedocs.io/en/stable/data-frame.md
 
 ## Development
 
-To install a virtualenv for development, use
-
-    make venv
-
-or, for a specific Python version
+Install uv from https://docs.astral.sh/uv/ and then run
 
-    make PYTHON=python3.X venv
+    uv sync

{dataiter-1.4 → dataiter-2.1}/dataiter/__init__.py

@@ -20,20 +20,32 @@
 # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 # THE SOFTWARE.
 
-import contextlib
-import numpy as np
-
-from dataiter import util
-
-__version__ = "1.4"
+__version__ = "2.1"
 
+#: Default amount of elements to return from :meth:`Vector.head`,
+#: :meth:`Vector.sample` and :meth:`Vector.tail`.
 DEFAULT_PEEK_ELEMENTS = 10
+
+#: Default amount of items to return from :meth:`ListOfDicts.head`,
+#: :meth:`ListOfDicts.sample` and :meth:`ListOfDicts.tail`.
 DEFAULT_PEEK_ITEMS = 3
+
+#: Default amount of rows to return from :meth:`DataFrame.head`,
+#: :meth:`DataFrame.sample` and :meth:`DataFrame.tail`.
 DEFAULT_PEEK_ROWS = 10
+
+#: Amount of significant digits to print floats with, all values of a
+#: column or vector using the decimals that the largest value gets.
 PRINT_FLOAT_PRECISION = 6
+
+#: Maximum amount of elements to show when printing a :class:`Vector`.
 PRINT_MAX_ELEMENTS = 100
+
+#: Maximum amount of items to show when printing a :class:`ListOfDicts`.
 PRINT_MAX_ITEMS = 10
-PRINT_MAX_ROWS = 100
+
+#: Maximum amount of rows to show when printing a :class:`DataFrame`.
+PRINT_MAX_ROWS = 20
 
 #: Maximum amount of columns to wrap print output to. This is only a fallback
 #: in case Python's ``shutil.get_terminal_size`` fails to detect the width of
@@ -49,46 +61,9 @@ PRINT_THOUSAND_SEPARATOR = ""
 #: rendered at the cut point.
 PRINT_TRUNCATE_WIDTH = 36
 
-#: ``True`` to use Numba, if available, to speed up :doc:`aggregations
-#: </aggregation>`, ``False`` to only use pure Python code.
-USE_NUMBA = False
-
-#: ``True`` to use Numba cache for JIT-compiled :doc:`aggregations
-#: </aggregation>`, ``False`` to only keep compiled code in memory for the
-#: duration of the session.
-USE_NUMBA_CACHE = True
-
-if not np.__version__.startswith("2."):
-    raise Exception("NumPy 2.x required")
-
-with contextlib.suppress(LookupError):
-    USE_NUMBA_CACHE = util.parse_env_boolean("DATAITER_USE_NUMBA_CACHE")
-
-try:
-    # Force Numba on or off if environment variable defined.
-    USE_NUMBA = util.parse_env_boolean("DATAITER_USE_NUMBA")
-except LookupError:
-    with contextlib.suppress(Exception):
-        # Use Numba automatically if found
-        # and calling a trivial function works.
-        import numba
-        try:
-            @numba.njit(cache=USE_NUMBA_CACHE)
-            def check(x):
-                return x**2
-            assert check(10) == 100
-            USE_NUMBA = True
-        except Exception as error:
-            print(f"Numba found, but disabled due to error: {error!s}")
-
-globals().pop("check", None)
-globals().pop("contextlib", None)
-globals().pop("np", None)
-globals().pop("numba", None)
-globals().pop("util", None)
-
 from dataiter import dtypes
 from dataiter.vector import Vector
+from dataiter.vector import of_type
 from dataiter.data_frame import DataFrame
 from dataiter.data_frame import DataFrameColumn
 from dataiter.geojson import GeoJSON
@@ -108,13 +83,35 @@ from dataiter.aggregate import median
 from dataiter.aggregate import min
 from dataiter.aggregate import mode
 from dataiter.aggregate import nth
+from dataiter.aggregate import prod
 from dataiter.aggregate import quantile
 from dataiter.aggregate import std
 from dataiter.aggregate import sum
 from dataiter.aggregate import var
 
+from dataiter.window import cummax
+from dataiter.window import cummin
+from dataiter.window import cumsum
+from dataiter.window import rank
+from dataiter.window import shift
+
 from dataiter.io import read_csv
+from dataiter.io import read_duckdb
 from dataiter.io import read_geojson
 from dataiter.io import read_json
+from dataiter.io import read_jsonl
 from dataiter.io import read_npz
 from dataiter.io import read_parquet
+from dataiter.io import read_parquet_chunked
+
+from dataiter import util
+
+#: Special value to refer to the next element/item.
+#: Used by ``replace_na`` to replace missing values with
+#: the next non-missing value, i.e. to fill backward.
+NEXT = util.Sentinel("NEXT")
+
+#: Special value to refer to the previous element/item.
+#: Used by ``replace_na`` to replace missing values with
+#: the previous non-missing value, i.e. to fill forward.
+PREVIOUS = util.Sentinel("PREVIOUS")