Struct SourceDocument 

pub struct SourceDocument {
    pub name: String,
    pub format: InputFormat,
    pub bytes: Vec<u8>,
    pub path: Option<PathBuf>,
    pub base_url: Option<String>,
    pub encoding: Option<String>,
}

A loaded input document: its name, detected format, and raw bytes.

Fields

name: String

format: InputFormat

bytes: Vec<u8>

path: Option<PathBuf>

The filesystem path it was loaded from, if any (from_file). Used to resolve relative <img src> paths when image fetching is enabled; None for in-memory sources.

base_url: Option<String>

The URL this document was fetched from, if any. Used to resolve relative / protocol-relative <img src> against the page’s origin when image fetching is enabled (an HTML page fetched from the web references its images by relative path). None for local / in-memory sources.

encoding: Option<String>

The character encoding to decode the bytes with when a backend reads them as text — docling’s TextBackendOptions.encoding (MarkdownBackendOptions(encoding="shift_jis")). A WHATWG encoding label (shift_jis, koi8-r, windows-1251, latin1; Python codec spellings with _ are accepted too). None (default) detects the encoding — see Self::text. Set by Self::with_encoding or, for every source a converter handles, DocumentConverter::encoding.

Implementations

impl SourceDocument

pub fn from_file(path: impl AsRef<Path>) -> Result<Self, ConversionError>

Load a document from a filesystem path, detecting the format from the extension.

pub fn from_bytes( name: impl Into<String>, format: InputFormat, bytes: Vec<u8>, ) -> Self

Construct directly from in-memory bytes (no disk access).

pub fn with_encoding(self, label: Option<String>) -> Self

Decode the bytes with this character encoding when read as text — nothing is guessed, and bytes the encoding cannot decode fail the conversion (docling raises DocumentLoadError). None restores detection.

pub fn with_base_url(self, url: impl Into<String>) -> Self

Record the URL this document was fetched from (for resolving relative <img src> against the page origin when image fetching is enabled).

pub fn base_dir(&self) -> Option<&Path>

The directory containing the source file, for resolving relative asset paths. None for in-memory sources.

pub fn text(&self) -> Result<Cow<'_, str>, ConversionError>

The bytes as text, for text-based backends — docling’s decode_text (docling#4202, 2.129). A byte-order mark settles the encoding first (UTF-32, UTF-8, UTF-16, and is dropped — docling#4098/#4109: Excel and Google Sheets write one when exporting “CSV UTF-8”, and kept it prefixes the first cell/line, so # Title stops being a heading and WebVTT’s signature check misses WEBVTT). Without one the text must be UTF-8; when it is not, the share of non-ASCII bytes that still form well-formed UTF-8 sequences decides: at 50% or more the file reads as damaged UTF-8 and is an error (decoding it as a code page would turn the damage into text that looks fine and is not), below that it is the one code page docling falls back to, windows-1252 — decoded with a warning, or an error on the five bytes that page leaves undefined (Python’s cp1252 rejects them where the WHATWG table maps them to C1 controls). Borrowed when the bytes are already UTF-8.

Trait Implementations

impl Clone for SourceDocument

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SourceDocument

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.