felipecrv commented on code in PR #41593: URL: https://github.com/apache/arrow/pull/41593#discussion_r1637342497
########## docs/source/format/Intro.rst: ########## @@ -0,0 +1,485 @@ +.. Licensed to the Apache Software Foundation (ASF) under one +.. or more contributor license agreements. See the NOTICE file +.. distributed with this work for additional information +.. regarding copyright ownership. The ASF licenses this file +.. to you under the Apache License, Version 2.0 (the +.. "License"); you may not use this file except in compliance +.. with the License. You may obtain a copy of the License at + +.. http://www.apache.org/licenses/LICENSE-2.0 + +.. Unless required by applicable law or agreed to in writing, +.. software distributed under the License is distributed on an +.. "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +.. KIND, either express or implied. See the License for the +.. specific language governing permissions and limitations +.. under the License. + +************ +Introduction +************ + +Apache Arrow was born from the need for a set of standards around +tabular data representation and interchange between systems. +The adoption of these standards reduce computing costs of data +serialization/deserialization and implementation costs across +systems implemented in different programming languages. + +The Apache Arrow specification can be implemented in any programming +language but official implementations for many languages are available. +An implementation consists of format definitions using the constructs +offered by the language and common in-memory data processing algorithms +(e.g. slicing and concatenating). Users can extend and use the utilities +provided by the Apache Arrow implementation in their programming +language of choice. Some implementations are further ahead and feature a +vast set of algorithms for in-memory analytical data processing. + +As the format gets more adoption, it becomes easier for data processing +systems to exchange tabular data. Among other things, an agreed upon +in-memory format, enables the implementations of zero-copy IPC protocols +(inter-process communication without copying data in memory) and +more efficient reading and writing of file formats like CSV, `Apache ORC`_, +and `Apache Parquet`_. + +.. _Apache ORC: https://orc.apache.org/ +.. _Apache Parquet: https://parquet.apache.org/ + +Arrow Columnar Format +===================== + +Apache Arrow focuses on tabular data so let's consider we have data +which are tabular so they can be organized into a table: + +.. figure:: ./images/columnar-diagram_1.svg + :scale: 70% + :alt: Diagram with tabular data of 4 rows and columns. + +This kind of data can be represented in memory using a row based format or a +column based format. The row based format stores data by row meaning the rows +are adjacent in the computer memory: + +.. figure:: ./images/columnar-diagram_2.svg + :alt: Tabular data being structured row by row in computer memory. + +In a columnar format, on the other hand, the data is organized by column +instead of by row making analytical operations like filtering, grouping, +aggregations and others more efficient because the CPU can maintain memory locality +and require less memory jumps to process the data. By keeping the data contiguous +in memory it also enables vectorization of the computations. Most modern +CPUs have single instructions, multiple data (SIMD) enabling parallel +processing and execution of operations on vector data using a single CPU +instruction. + +Apache Arrow is solving this exact problem. It is the specification that +uses the columnar layout. + +.. figure:: ./images/columnar-diagram_3.svg + :alt: Tabular data being structured column by column in computer memory. + +The column is called an **Array** in Arrow terminology. Arrays can be of +different data types and the way their values are stored in memory varies among +the data types. The specification of how these values are arranged in memory is +what we call a **physical memory layout**. One contiguous region of memory that +stores data for arrays is called a **Buffer**. + +Next sections give an introduction to Arrow Columnar Format explaining the +different physical layouts. The full specification of the format can be found +at :ref:`format_columnar`. + +Support for null values +======================= + +Arrow supports missing values or "nulls" for all data types: any value +in an array may be semantically null, whether primitive or nested data type. + +In Arrow, a dedicated buffer, known as the validity (or "null") bitmap, +is used alongside the data indicating whether each value in the array is +null or not: a value of 1 means that the value is not-null ("valid"), whereas +a value of 0 indicates that the value is null. + +This validity bitmap is optional: if there are no missing values in +the array the buffer does not need to be allocated (as in the example +column 1 in the diagram below). + +.. note:: + + We read validity bitmaps right-to-left within a group of 8 bits due to + `bit-endianness <https://en.wikipedia.org/wiki/Bit_numbering>`_ being + used. + +Primitive layouts +================= + +Fixed Size Primitive Layout +--------------------------- + +A primitive column represents an array of values where each value +has the same physical size measured in bytes. Data types that use the +fixed size primitive layout are, for example, signed and unsigned +integer data types, floating point numbers, boolean, decimal and temporal +data types. + +.. figure:: ./images/primitive-diagram.svg + :alt: Diagram is showing the difference between the primitive data + type presented in a Table and the data actually stored in + computer memory. + + Physical layout diagram for primitive data types. + +.. note:: + Boolean data type is represented with a primitive layout where the + values are encoded in bits instead of bytes. That means the physical + layout includes a values bitmap buffer and possibly a validity bitmap + buffer. + + .. figure:: ./images/bool-diagram.svg + :alt: Diagram is showing the difference between the boolean data + type presented in a Table and the data actually stored in + computer memory. + + Physical layout diagram for boolean data type. + +.. note:: + Arrow also has a concept of Null data type where all values are null. In + this case no buffers are allocated. + +Variable length binary and string +--------------------------------- + +The bytes of all elements in a binary or string column are stored together +consecutively in a single buffer or region of memory. To know where each element +of the column starts and ends the physical layout also includes integer offsets. +The number of elements of the offset buffer is one more than the length of the +array as the last two elements define the start and the end of the last +element in the binary/string column. + +Binary and string data types share the same physical layout. The one difference +between them is that the string data type is utf-8 binary and assumes to contain +utf-8 encoded strings. + +The difference between binary/string and large binary/string is in the offset +data type. In the first case that is int32 and in the second it is int64. + +The limitation of data types using 32 bit offsets is that they have a max size of +2GB per array. One can still use the non-large variants for bigger data, but +then multiple chunks are needed. + +.. figure:: ./images/var-string-diagram.svg + :alt: Diagram is showing the difference between the variable length + string data type presented in a Table and the data actually + stored in computer memory. + + Physical layout diagram for variable length string data types. + +Variable length binary and string view Review Comment: ```suggestion Variable-Length Binary and String View ``` ########## docs/source/format/Intro.rst: ########## @@ -0,0 +1,485 @@ +.. Licensed to the Apache Software Foundation (ASF) under one +.. or more contributor license agreements. See the NOTICE file +.. distributed with this work for additional information +.. regarding copyright ownership. The ASF licenses this file +.. to you under the Apache License, Version 2.0 (the +.. "License"); you may not use this file except in compliance +.. with the License. You may obtain a copy of the License at + +.. http://www.apache.org/licenses/LICENSE-2.0 + +.. Unless required by applicable law or agreed to in writing, +.. software distributed under the License is distributed on an +.. "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +.. KIND, either express or implied. See the License for the +.. specific language governing permissions and limitations +.. under the License. + +************ +Introduction +************ + +Apache Arrow was born from the need for a set of standards around +tabular data representation and interchange between systems. +The adoption of these standards reduce computing costs of data +serialization/deserialization and implementation costs across +systems implemented in different programming languages. + +The Apache Arrow specification can be implemented in any programming +language but official implementations for many languages are available. +An implementation consists of format definitions using the constructs +offered by the language and common in-memory data processing algorithms +(e.g. slicing and concatenating). Users can extend and use the utilities +provided by the Apache Arrow implementation in their programming +language of choice. Some implementations are further ahead and feature a +vast set of algorithms for in-memory analytical data processing. + +As the format gets more adoption, it becomes easier for data processing +systems to exchange tabular data. Among other things, an agreed upon +in-memory format, enables the implementations of zero-copy IPC protocols +(inter-process communication without copying data in memory) and +more efficient reading and writing of file formats like CSV, `Apache ORC`_, +and `Apache Parquet`_. + +.. _Apache ORC: https://orc.apache.org/ +.. _Apache Parquet: https://parquet.apache.org/ + +Arrow Columnar Format +===================== + +Apache Arrow focuses on tabular data so let's consider we have data +which are tabular so they can be organized into a table: + +.. figure:: ./images/columnar-diagram_1.svg + :scale: 70% + :alt: Diagram with tabular data of 4 rows and columns. + +This kind of data can be represented in memory using a row based format or a +column based format. The row based format stores data by row meaning the rows +are adjacent in the computer memory: + +.. figure:: ./images/columnar-diagram_2.svg + :alt: Tabular data being structured row by row in computer memory. + +In a columnar format, on the other hand, the data is organized by column +instead of by row making analytical operations like filtering, grouping, +aggregations and others more efficient because the CPU can maintain memory locality +and require less memory jumps to process the data. By keeping the data contiguous +in memory it also enables vectorization of the computations. Most modern +CPUs have single instructions, multiple data (SIMD) enabling parallel +processing and execution of operations on vector data using a single CPU +instruction. + +Apache Arrow is solving this exact problem. It is the specification that +uses the columnar layout. + +.. figure:: ./images/columnar-diagram_3.svg + :alt: Tabular data being structured column by column in computer memory. + +The column is called an **Array** in Arrow terminology. Arrays can be of +different data types and the way their values are stored in memory varies among +the data types. The specification of how these values are arranged in memory is +what we call a **physical memory layout**. One contiguous region of memory that +stores data for arrays is called a **Buffer**. + +Next sections give an introduction to Arrow Columnar Format explaining the +different physical layouts. The full specification of the format can be found +at :ref:`format_columnar`. + +Support for null values +======================= + +Arrow supports missing values or "nulls" for all data types: any value +in an array may be semantically null, whether primitive or nested data type. + +In Arrow, a dedicated buffer, known as the validity (or "null") bitmap, +is used alongside the data indicating whether each value in the array is +null or not: a value of 1 means that the value is not-null ("valid"), whereas +a value of 0 indicates that the value is null. + +This validity bitmap is optional: if there are no missing values in +the array the buffer does not need to be allocated (as in the example +column 1 in the diagram below). + +.. note:: + + We read validity bitmaps right-to-left within a group of 8 bits due to + `bit-endianness <https://en.wikipedia.org/wiki/Bit_numbering>`_ being + used. + +Primitive layouts +================= + +Fixed Size Primitive Layout +--------------------------- + +A primitive column represents an array of values where each value +has the same physical size measured in bytes. Data types that use the +fixed size primitive layout are, for example, signed and unsigned +integer data types, floating point numbers, boolean, decimal and temporal +data types. + +.. figure:: ./images/primitive-diagram.svg + :alt: Diagram is showing the difference between the primitive data + type presented in a Table and the data actually stored in + computer memory. + + Physical layout diagram for primitive data types. + +.. note:: + Boolean data type is represented with a primitive layout where the + values are encoded in bits instead of bytes. That means the physical + layout includes a values bitmap buffer and possibly a validity bitmap + buffer. + + .. figure:: ./images/bool-diagram.svg + :alt: Diagram is showing the difference between the boolean data + type presented in a Table and the data actually stored in + computer memory. + + Physical layout diagram for boolean data type. + +.. note:: + Arrow also has a concept of Null data type where all values are null. In + this case no buffers are allocated. + +Variable length binary and string Review Comment: ```suggestion Variable-Length Binary and String ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
