Module @pdfrx/viewer-core

@pdfrx/viewer-core — the platform-independent core logic of the pdfrx_web viewer.

This package contains no DOM access: it is pure geometry, layout, viewport math, text-flow analysis, and text-selection logic. All public types are plain, JSON-serializable objects so the exact same test vectors can be shared across implementations.

Modules:

geometry — rect/point math, rotation, and PDF page space <-> document space conversions.
transform — the ViewTransform (uniform scale + translation), fit calculations, page anchors, and boundary/overscroll clamping.
layout — vertical/horizontal page layout.
text / text-formatter — the structured page text model and its flow analysis (reading order, line splitting, word/space/newline fragments).
selection — nearest-character hit testing, A/B anchors, word selection, and per-page range expansion.
page-arrangement — stable page-placement identity and pure operations for application-owned editing and synchronization models.

Coordinate spaces: PDF page space is points (1/72"), origin bottom-left, y-up, top >= bottom; document space is y-down over the whole laid-out document; view space is document space transformed by a ViewTransform.

@pdfrx/viewer-core

Platform-independent core logic of a PDF viewer, in pure TypeScript. No DOM access — every type is a plain JSON-serializable object. This is the logic layer underneath @pdfrx/viewer; use it directly to build a custom viewer shell (different rendering stack, custom gestures).

_{Derived from the pdfrx project.}

npm package · API reference

Installation

npm install @pdfrx/viewer-core

What's inside

Module	Contents
`geometry`	Rect/point math, page rotation, PDF page space (y-up) ↔ document space (y-down) conversions
`transform`	`ViewTransform {zoom, xZoomed, yZoomed}`, visible-rect/fit calculations, 14 page anchors, boundary clamping, underflow alignment
`layout`	Vertical/horizontal page layout and hit testing
`text` / `text-formatter`	Structured page text: reading-order analysis, line splitting, word/space/newline fragments, text direction (LTR/RTL/vertical), search
`selection`	Nearest-character hit testing, A/B selection anchors (same-page and cross-page), word selection, per-page range expansion for highlighting

These primitives are DOM-free and JSON-serializable. They support vertical, horizontal, and custom viewer shells without coupling layout, coordinates, text analysis, or selection behavior to canvas rendering or a UI framework.

Example

import {
  formatText,
  findTextAndIndexForPoint,
  getSelectedRanges,
  composeSelectedText,
  layoutPagesVertical,
} from '@pdfrx/viewer-core';

// Build structured text from an engine's raw page text
const pageText = formatText({ fullText, charRects }, /* pageNumber */ 1);

// Lay out pages and hit-test a pointer position against the text
const layout = layoutPagesVertical(pages, { margin: 8 });
const hit = findTextAndIndexForPoint(docPoint, selectablePages);

// Expand a selection into per-page ranges and compose the clipboard text
const ranges = getSelectedRanges(selA, selB, getLoadedText);
const clipboard = composeSelectedText(ranges);

Coordinate conventions: PDF page space is points (1/72"), origin bottom-left, y-up; document space is y-down. See the architecture notes for details.

Next steps

See the complete API reference for exported geometry, layout, text, and selection symbols. The architecture notes define coordinate conventions and package boundaries.