mirror of
https://github.com/bulwarkmail/webmail.git
synced 2026-11-04 00:14:12 +00:00
Linus Rath
7c97eb2691
Every reader now stops at the read's character budget, a cell at 4,000 characters, one Office XML part at 12 MB, and parts with a DTD are refused. A compound file named .docx is encrypted only when it holds an EncryptedPackage stream, else unsupported; .eml files are unsupported instead of raw MIME; maxChars that isn't a number falls back to 20,000. attachment-text.test.ts: the zip bomb test passes the cap as an option now, and a bare compound file named secret.docx is expected unsupported (it is a legacy file under a new name; limits.test.ts covers the password-protected one).
61 lines
2.5 KiB
TypeScript
61 lines
2.5 KiB
TypeScript
// Shapes and limits of the text extraction for attachments and files.
|
|
// Pure: no imports from the app, so the plugin host and Files can share it.
|
|
|
|
/** What kind of file the text came from. */
|
|
export type TextFormat =
|
|
| 'text' | 'csv' | 'html' | 'ics' | 'vcf' | 'pdf'
|
|
| 'docx' | 'xlsx' | 'pptx' | 'odt' | 'ods' | 'odp' | 'other';
|
|
|
|
/**
|
|
* ok: `text` holds the text. too_big: over the size limit, not read.
|
|
* unsupported: a format without a reader (images, legacy .doc/.xls/.ppt,
|
|
* archives). no_text: nothing to read, e.g. a scanned PDF. encrypted:
|
|
* protected by a password. failed: the file is damaged or reading broke.
|
|
*/
|
|
export type TextStatus = 'ok' | 'too_big' | 'unsupported' | 'no_text' | 'encrypted' | 'failed';
|
|
|
|
export interface ExtractedText {
|
|
format: TextFormat;
|
|
status: TextStatus;
|
|
/** The text, '' unless status is ok. */
|
|
text: string;
|
|
/** True when the text was cut at maxChars, or a PDF or sheet had more than was read. */
|
|
truncated: boolean;
|
|
/** PDFs: the number of pages in the file. */
|
|
pages?: number;
|
|
/** Spreadsheets: the names of the sheets read. */
|
|
sheets?: string[];
|
|
}
|
|
|
|
export interface ExtractOptions {
|
|
/** File name; its extension picks the format first. */
|
|
name?: string;
|
|
/** Media type, used when the name says nothing. */
|
|
type?: string;
|
|
/** Longest text returned (default DEFAULT_TEXT_CHARS, at most MAX_TEXT_CHARS). */
|
|
maxChars?: number;
|
|
}
|
|
|
|
/** Text returned when the caller doesn't ask for a length. */
|
|
export const DEFAULT_TEXT_CHARS = 20_000;
|
|
/** Longest text one read returns. */
|
|
export const MAX_TEXT_CHARS = 100_000;
|
|
/** Largest file read for its text. */
|
|
export const MAX_TEXT_SOURCE_BYTES = 20 * 1024 * 1024;
|
|
/** Pages of a PDF read at most. */
|
|
export const MAX_PDF_PAGES = 50;
|
|
/** Fewer characters per page than this on average: the PDF has no text layer. */
|
|
export const MIN_PDF_CHARS_PER_PAGE = 20;
|
|
/** Sheets of a spreadsheet read at most. */
|
|
export const MAX_SHEETS = 5;
|
|
/** Rows per sheet read at most. */
|
|
export const MAX_SHEET_ROWS = 2_000;
|
|
/** Bytes an Office file may unpack to, all entries read together (zip bombs). */
|
|
export const MAX_INFLATED_BYTES = 50 * 1024 * 1024;
|
|
/** Bytes one XML part of an Office file may unpack to: DOMParser holds it all in memory. */
|
|
export const MAX_XML_PART_BYTES = 12 * 1024 * 1024;
|
|
/** Characters of one spreadsheet cell, after which it is cut. */
|
|
export const MAX_CELL_CHARS = 4_000;
|
|
/** Readers stop this many characters past maxChars, so the final cut can end at a word. */
|
|
export const TEXT_BUDGET_SLACK = 2_000;
|