oconvPdfFontDecoder

Per-font character decoder for the tier-1 pdf reader — code → Unicode text, and code → glyph width.

Module oconvPdfFontDecoder (oconvPdfFontDecoder) | Source packages/front/office/oconv/src/read/pdf/font-decoder.js | Deps pdfFont, pdfFontEncoding, pdfFilterDispatch, cmapToUnicode, encodingLookup, encodingAgl, standard14Lookup | Worker-safe yes

Turns a PDF Font dict into a decoder honouring the tier-1 decode paths in priority order: a /ToUnicode CMap (authoritative), then an /Encoding table + the AGL glyph-name hop for simple fonts, then — for composite (Type0) fonts only — the predefined CMap the /Encoding names, when that CMap's input codes are Unicode. The same decoder also gives each code's glyph width, which oconvPdfTextExtract uses to place the end of every text piece. Composes only documented public exports of @awacloud/pdf and @awacloud/fonts — no internal reach.

Resolve

import { ModuleRuntime } from '@awacloud/fw/core/runtime.js';
import { fw_require, modules } from '@awacloud/oconv';

const runtime = new ModuleRuntime();
runtime.registerAll(fw_require);
runtime.registerAll(modules);
const { buildDecoder, decodeShow } = runtime.resolve('oconvPdfFontDecoder');

API

Examples

Decode a show-operator string through a ToUnicode CMap

const pdfParserObj = runtime.resolve('pdfParser').obj;
const cmapToUnicode = runtime.resolve('cmapToUnicode');
const enc = (s) => new TextEncoder().encode(s);

const toUniSrc = cmapToUnicode.buildToUnicode(new Map([[0x41, 'A'], [0x42, 'B']]));
const fontDict = pdfParserObj.dict({
    Type: pdfParserObj.name('Font'), Subtype: pdfParserObj.name('Type1'),
    BaseFont: pdfParserObj.name('Helvetica'),
    ToUnicode: pdfParserObj.stream(pdfParserObj.dict({}), enc(toUniSrc))
});

const decoder = buildDecoder(fontDict, (ref) => ref);
decodeShow(decoder, new Uint8Array([0x41, 0x42]));
// { text: 'AB', decoded: 2, undecodable: 0 }

Notes

• Decode priority: /ToUnicode first; when absent (or it has no entry for a code), the /Encoding table + encodingAgl.glyphNameToUnicode hop is tried for simple (non-Type0) fonts only.
• Standard CMap: a Type0 font whose /Encoding names a Uni{GB,CNS,JIS,KS}-{UCS2,UTF16}[-HW]-{H,V} predefined CMap (ISO 32000-2 §9.7.5.2) decodes a code without a ToUnicode entry as the UTF-16 code unit it is; a surrogate half is counted undecodable. Identity-H/Identity-V (codes are CIDs, not Unicode) and the legacy RKSJ/EUC/… CMaps (tables not bundled) decode nothing here — a composite font in that position with no ToUnicode stays an honest text/undecodable loss (CID→GID→Unicode font-program walking is out of tier-1 scope).
• A code neither path resolves is counted undecodable and omitted from the text — never silently kept as a wrong glyph, never dropped without a count (decodeShow's own return shape).
• Glyph widths (width), by font kind:
  • simple fonts: /Widths indexed from /FirstChar. A code outside the array takes the descriptor's /MissingWidth (0 when absent, the spec default). A Type 3 font's widths are scaled by its /FontMatrix.
  • composite (Type0) fonts: the descendant CIDFont's /W array, in both forms (c [w1 w2 …] and cFirst cLast w), with /DW (default 1000) for CIDs it does not list. Codes are read as CIDs only under Identity-H/Identity-V. Under any other CMap the code → CID map is unknown here, so /DW is used.
  • Standard 14 fonts with no /Widths: the Adobe Core 14 AFM widths shipped by @awacloud/fonts (standard14Lookup), looked up by the glyph name the font's encoding gives the code. Symbol and ZapfDingbats are looked up by code.
• When a width cannot be read from the font, a declared fallback is used: 500 thousandths for a simple font, 1000 for a composite one. widthApproximated() then turns true, and oconvPdfTextExtract records text/width-approximated. The cases are: /Widths absent on a font that is not one of the Standard 14, a malformed width array or entry, a missing descendant font, a /W table under a non-identity CMap, and a named Standard 14 glyph missing from the AFM table. A malformed width structure never throws.
• cidBytes is 2 for Type0 fonts, 1 otherwise — decodeShow uses it to split strBytes into codes.

See also

• docs/api/README.md — full module index + exports boundary note
• oconvPdfTextExtract — the sole consumer of this decoder
• oconvPdfToIr — the reader composing the pdf pipeline
• loss matrix