vortex_file/lib.rs
1// SPDX-License-Identifier: Apache-2.0
2// SPDX-FileCopyrightText: Copyright the Vortex contributors
3
4#![allow(clippy::cast_possible_truncation)]
5#![doc(html_logo_url = "/vortex/docs/_static/vortex_spiral_logo.svg")]
6//! Read and write Vortex layouts, a serialization of Vortex arrays.
7//!
8//! A layout is a serialized array which is stored in some linear and contiguous block of
9//! memory. Layouts are recursive, and there are currently three types:
10//!
11//! 1. The [`FlatLayout`](vortex_layout::layouts::flat::FlatLayout). A contiguously serialized array of buffers, with a specific in-memory [`Alignment`](vortex_buffer::Alignment).
12//!
13//! 2. The [`StructLayout`](vortex_layout::layouts::struct_::StructLayout). Each column of a
14//! [`StructArray`][vortex_array::arrays::StructArray] is sequentially laid out at known offsets.
15//! This permits reading a subset of columns in linear time, as well as constant-time random
16//! access to any column.
17//!
18//! 3. The [`ChunkedLayout`](vortex_layout::layouts::chunked::ChunkedLayout). Each chunk of a
19//! [`ChunkedArray`](vortex_array::arrays::ChunkedArray) is sequentially laid out at known
20//! offsets. Finding the chunks containing row range is an `Nlog(N)` operation of searching the
21//! offsets.
22//!
23//! 4. The [`ZonedLayout`](vortex_layout::layouts::zoned::ZonedLayout).
24//!
25//! A layout, alone, is _not_ a standalone Vortex file because layouts are not self-describing. They
26//! neither contain a description of the kind of layout (e.g. flat, column of flat, chunked of
27//! column of flat) nor a data type ([`DType`](vortex_dtype::DType)).
28//!
29//! # Reading
30//!
31//! Vortex files are read using [`VortexOpenOptions`], which can be provided with information about the file's
32//! structure to save on IO before the actual data read. Once the file is open and has done the initial IO work to understand its own structure,
33//! it can be turned into a stream by calling [`VortexFile::scan`].
34//!
35//! The file manages IO-oriented work and CPU-oriented work on two different underlying runtimes, which are configurable and pluggable with multiple provided implementations (Tokio, Rayon etc.).
36//!
37//! # File Format
38//!
39//! Succinctly, the file format specification is as follows:
40//!
41//! 1. Data is written first, in a form that is describable by a Layout (typically Array IPC Messages).
42//! 1. To allow for more efficient IO & pruning, our writer implementation first writes the "data" arrays,
43//! and then writes the "metadata" arrays (i.e., per-column statistics)
44//! 2. We write what is collectively referred to as the "Footer", which contains:
45//! 1. An optional Schema, which if present is a valid flatbuffer representing a message::Schema
46//! 2. The Layout, which is a valid footer::Layout flatbuffer, and describes the physical byte ranges & relationships amongst
47//! the those byte ranges that we wrote in part 1.
48//! 3. The Postscript, which is a valid footer::Postscript flatbuffer, containing the absolute start offsets of the Schema & Layout
49//! flatbuffers within the file.
50//! 4. The End-of-File marker, which is 8 bytes, and contains the u16 version, u16 postscript length, and 4 magic bytes.
51//!
52//! ## Illustrated File Format
53//! ```text
54//! ┌────────────────────────────┐
55//! │ │
56//! │ Data │
57//! │ (Array IPC Messages) │
58//! │ │
59//! ├────────────────────────────┤
60//! │ │
61//! │ Per-Column Statistics │
62//! │ │
63//! ├────────────────────────────┤
64//! │ │
65//! │ Schema Flatbuffer │
66//! │ │
67//! ├────────────────────────────┤
68//! │ │
69//! │ Layout Flatbuffer │
70//! │ │
71//! ├────────────────────────────┤
72//! │ │
73//! │ Postscript Flatbuffer │
74//! │ (Schema & Layout Offsets) │
75//! │ │
76//! ├────────────────────────────┤
77//! │ 8-byte End of File │
78//! │(Version, Postscript Length,│
79//! │ Magic Bytes) │
80//! └────────────────────────────┘
81//! ```
82//!
83//! A Parquet-style file format is realized by using a chunked layout containing column layouts
84//! containing chunked layouts containing flat layouts. The outer chunked layout represents row
85//! groups. The inner chunked layout represents pages.
86//!
87//! Layouts are adaptive, and the writer is free to build arbitrarily complex layouts to suit their
88//! goals of locality or parallelism. For example, one may write a column in a Struct Layout with
89//! or without chunking, or completely elide statistics to save space or if they are not needed, for
90//! example if the metadata is being stored in an external index.
91//!
92
93mod counting;
94mod file;
95mod footer;
96mod open;
97mod pruning;
98pub mod segments;
99mod strategy;
100#[cfg(test)]
101mod tests;
102mod writer;
103
104pub use file::*;
105pub use footer::*;
106pub use forever_constant::*;
107pub use open::*;
108pub use strategy::*;
109use vortex_alp::{ALPEncoding, ALPRDEncoding};
110use vortex_array::arrays::DictEncoding;
111use vortex_array::{ArraySessionExt, EncodingRef};
112use vortex_bytebool::ByteBoolEncoding;
113use vortex_datetime_parts::DateTimePartsEncoding;
114use vortex_decimal_byte_parts::DecimalBytePartsEncoding;
115use vortex_fastlanes::{BitPackedEncoding, DeltaEncoding, FoREncoding, RLEEncoding};
116use vortex_fsst::FSSTEncoding;
117use vortex_pco::PcoEncoding;
118use vortex_runend::RunEndEncoding;
119use vortex_sequence::SequenceEncoding;
120use vortex_session::VortexSession;
121use vortex_sparse::SparseEncoding;
122use vortex_zigzag::ZigZagEncoding;
123pub use writer::*;
124
125/// The current version of the Vortex file format
126pub const VERSION: u16 = 1;
127/// The size of the footer in bytes in Vortex version 1
128pub const V1_FOOTER_FBS_SIZE: usize = 32;
129
130/// Constants that will never change (i.e., doing so would break backwards compatibility)
131mod forever_constant {
132 /// The extension for Vortex files
133 pub const VORTEX_FILE_EXTENSION: &str = "vortex";
134
135 /// The maximum length of a Vortex postscript in bytes
136 pub const MAX_POSTSCRIPT_SIZE: u16 = u16::MAX - 8;
137 /// The magic bytes for a Vortex file
138 pub const MAGIC_BYTES: [u8; 4] = *b"VTXF";
139 /// The size of the EOF marker in bytes
140 pub const EOF_SIZE: usize = 8;
141
142 #[cfg(test)]
143 mod test {
144 use super::*;
145 use crate::*;
146
147 #[test]
148 fn never_change_these_constants() {
149 assert_eq!(V1_FOOTER_FBS_SIZE, 32);
150 assert_eq!(MAX_POSTSCRIPT_SIZE, 65527);
151 assert_eq!(MAGIC_BYTES, *b"VTXF");
152 assert_eq!(EOF_SIZE, 8);
153 }
154 }
155}
156
157/// Register the default encodings use in Vortex files with the provided session.
158///
159/// NOTE: this function will be changed in the future to encapsulate logic for using different
160/// Vortex "Editions" that may support different sets of encodings.
161pub fn register_default_encodings(session: &VortexSession) {
162 session.arrays().register_many([
163 EncodingRef::new_ref(ALPEncoding.as_ref()),
164 EncodingRef::new_ref(ALPRDEncoding.as_ref()),
165 EncodingRef::new_ref(BitPackedEncoding.as_ref()),
166 EncodingRef::new_ref(ByteBoolEncoding.as_ref()),
167 EncodingRef::new_ref(DateTimePartsEncoding.as_ref()),
168 EncodingRef::new_ref(DecimalBytePartsEncoding.as_ref()),
169 EncodingRef::new_ref(DeltaEncoding.as_ref()),
170 EncodingRef::new_ref(DictEncoding.as_ref()),
171 EncodingRef::new_ref(FSSTEncoding.as_ref()),
172 EncodingRef::new_ref(FoREncoding.as_ref()),
173 EncodingRef::new_ref(PcoEncoding.as_ref()),
174 EncodingRef::new_ref(RLEEncoding.as_ref()),
175 EncodingRef::new_ref(RunEndEncoding.as_ref()),
176 EncodingRef::new_ref(SequenceEncoding.as_ref()),
177 EncodingRef::new_ref(SparseEncoding.as_ref()),
178 EncodingRef::new_ref(ZigZagEncoding.as_ref()),
179 #[cfg(feature = "zstd")]
180 EncodingRef::new_ref(vortex_zstd::ZstdEncoding.as_ref()),
181 ]);
182}