Short: PDF datatypes - read PDFs on your Amiga Author: marcus.gerards@googlemail.com (Marcus Gerards) Uploader: marcus gerards googlemail com (Marcus Gerards) Type: util/dtype Version: 1.9 Replaces: util/dtype/PDFDatatype.lha Requires: AmigaOS 3.0 or newer. Builds for 68000/020/040/060 included. Architecture: m68k-amigaos >= 3.0 Distribution: NoCD Kurz: PDF-Datatypes: PDFs am Amiga lesen PDFDatatype 1.9 -- 28.8.2026 WHAT IT IS Three AmigaOS datatypes that teach the system to open PDF files. Install them once and a .pdf is an ordinary file: double-click it on Workbench, or open it in MultiView, and the page appears -- with buttons to turn pages and zoom, and text you can mark with the mouse and copy to the clipboard. Nothing else has to change. Any program that already shows a picture or a text file through the datatypes system now shows a PDF too. pdf.datatype plain text pages (a text.datatype subclass) pdfpic.datatype a page that is one image (a picture.datatype subclass) pdfdoc.datatype everything else (the general viewer) You do not choose between them; the file does. Each descriptor carries its own recognition code, reads the first page, and claims the file or declines it. Four descriptor files are installed for those three classes. The system sorts every file under a "text" or a "binary" heading by its first bytes and searches only that list, and PDFs fall on both sides depending on who wrote them -- so a fourth descriptor, PDFbin, waits on the binary side and hands those to pdfdoc. The manual explains it at length. WHAT IT READS Measured against real documents -- journal papers, slide decks, a hardware manual, contracts, a 290-page scanned book, work paperwork -- and not guessed from the specification. Yes PDF 1.0 to 1.7, from Acrobat, Word, LaTeX, OpenOffice, TCPDF, scanners, and from the Amiga's own PDFPrinter driver Yes compressed files (FlateDecode, RunLengthDecode), content compressed twice over (an ASCII85 wrapper round either), cross-reference streams, object streams, linearized files, page trees of any depth Yes positioned multi-font text, in colour, sideways, with umlauts and accents -- marked with the mouse and copied to the clipboard Yes scanned pages in black & white, grey and colour, JPEG included (handed to the system's own JPEG datatype) Yes figures placed on a text page Yes lines, boxes, rules, curves and filled shapes -- a schematic of a hundred thousand line segments draws in full No encrypted or password-protected files: refused with a message saying so No fillable forms and XFA, annotations, transparency, gradients and pattern fills, and re-usable drawing groups (Form XObjects) -- a page built entirely out of one says so in a framed note No page rotation and cropping: a page the document asks to be turned is shown upright, and the whole sheet is shown rather than the cropped part of it No images written inside the page's instructions (inline images) -- the picture is skipped, the text around it is not No CCITT G3/G4, JBIG2 and JPEG 2000 scans No embedded fonts: text is drawn in the nearest Amiga font at the nearest available size, spaced by the document's own measurements No non-Latin writing; no text search, no links, no editing The manual says all of this again, at length, in a form written for readers rather than for programmers. WHAT IT IS NOT A READING tool, not a proofing tool. It exists so a PDF can be read on a 640x256 screen, in sixteen colours, on a 68000 with two megabytes. It is NOT suitable for print production -- for judging, checking or approving anything going to a printing press. The typefaces are not the document's, the colours are the nearest your screen can manage, fine hairlines round to whole pixels, and some page content is not drawn at all and does not announce that it is missing. Use a real prepress application for that, on a machine built for it. For reading what a document says, what a drawing shows and what a table contains, this is exactly what it is for. WHAT YOU NEED AmigaOS 3.0 (V39) or newer -- that is when datatypes.library appeared. Developed and tested on AmigaOS 3.2. A 68000 with 2 MB is enough; big scans and dense drawings want more, and a page that will not fit is shown a zoom step smaller rather than refused. Pages that are pictures are decoded here and displayed by the system's own jpeg.datatype and bmp.datatype, both of which ship with AmigaOS. Text-only pages need neither. Four builds are supplied, one per processor family -- 68000, 68020, 68040 and 68060 -- and the Installer picks the one your machine has. About 400 KB installed. WHAT IS IN THE ARCHIVE Install the Installer script: processor, manual, languages, done Files/ four builds of the three classes, one per processor, and one set of descriptors shared by all of them (four files: three classes plus a catch-all for PDFs the system files as binary) Icons/ GlowIcons by Martin "Mason" Merz, offered by the Installer Docs/ the manual, in English and German, as AmigaGuide Tools/ DTInfo and DTLoad, Shell diagnostics for when a PDF opens as plain text INSTALLATION Double-click the Install icon. It asks which processor you have (already detected and pre-selected), whether to install the manual, and in which languages -- then copies the classes into SYS:Classes/DataTypes, the descriptors into DEVS:DataTypes, and registers them. No reboot. Then open a PDF. HISTORY 1.9 Added GlowIcons by Martin "Mason" Merz. The Installer offers them for the descriptors and the manual; the classic icons stay the default of the archive itself. 1.8 Black-and-white pages no longer come up blank: such a page was not drawn at all, while the grey and colour pages around it were. 1.7 The download shrinks by 65 KB: the second set of file-type descriptors added in 1.5 is reduced to one small catch-all, and updating removes the two files it replaces. 1.6 The description now lists two limits it had never mentioned: a page the document asks to be turned is shown upright, and a document asking for only part of the sheet gets the whole sheet. Both were always so; neither was written down. Inline images are named as something not drawn, and encrypted files are described as refused rather than as opening for gibberish. 1.5 PDFs that no datatype would open at all now open: recognition depended on which invisible marker bytes the producer wrote on line two, and half of them were never offered to this reader. Pages compressed a second time -- a chain of two filters, which ReportLab writes by default -- are read instead of refused. An image placed inside the page's own instructions no longer swallows the text around it. The image itself is still not drawn. An encrypted PDF says so instead of appearing to be damaged. 1.4 Words no longer run together: a word space squeezed by justification was read as kerning, so "and for" came out as "andfor". Letters no longer touch or melt into each other: sizes a bitmap font was never designed at are drawn from an outline font instead of a scaled bitmap, at every zoom. A word that comes out too wide is fitted into its own room instead of covering its neighbour, and spare room widens the word gaps instead of spreading the letters. Two fragments no longer print over each other when a document overstates its own text width. Copied text no longer doubles a space. 1.3 The three descriptors are built once and shared by every processor instead of one build each, which makes the download 16% smaller. 1.2 Ligatures no longer eat a letter -- "fixed" was appearing as "fxed" and "efforts" as "eforts", in every properly typeset document. A page whose whole content is a reusable drawing group -- usually a title page -- now says so in a framed note instead of coming up blank. 1.1 Very large documents open: books of sixty megabytes and up used to refuse to open at all, and every page of them is reachable now. Text written by modern word processors appears in full, where a recent Word document often showed only its capital letters with gaps where the rest of the words belonged. Apostrophes, quotation marks and ampersands were dropped the same way, so "What's" came out as "Whats"; they are back. Some pages that came up blank now show their picture -- two different ways of writing a page each made it vanish. A scan in a compression we cannot read now says so in a frame where the picture belongs, naming the compression, instead of being drawn as noise. And most such books can now be READ: where the scanner left its recognised text behind, that text is drawn instead, which turned one 980-page manual from white paper into a readable book. Background bands sit behind their pictures again instead of being painted over them, and a drop capital is fitted to the room the page gives it rather than drawn across the paragraph beside it. The installer's closing pages and its "needs AmigaOS 3.0" warning no longer break off mid-sentence, the Aminet description now travels inside the archive, and the drawers open showing their icons. 1.0 The initial version. CREDITS GlowIcons by Martin "Mason" Merz. Each one shipped is this project's own classic icon with his colour imagery appended, so it shows his artwork on AmigaOS 3.5 and newer and the plain icon on everything older. LICENCE Copyright (C) 2026 Marcus Gerards. All rights reserved. This covers the datatype classes, their descriptors, the diagnostic programs and their documentation, icons and other supplied files. You may download, install and use the Software free of charge for your own private and personal purposes. It may NOT be sold or otherwise distributed commercially, in whole or in part. It may NOT be bundled or combined with any other software, hardware, magazine, cover disk, collection or product. It may NOT otherwise be published, redistributed or mirrored, in any form, whether modified or unmodified. Each of these requires the author's prior written permission -- please ask first: Marcus Gerards marcus.gerards@googlemail.com The Software is provided "as is", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose and noninfringement. In no event shall the author be liable for any claim, damages or other liability arising from, out of or in connection with the Software or its use. All rights not expressly granted above are reserved by the author. The full text is in the manual, under "License".