Inside Type & Set

Type & Set is a typesetting system consisting of TEX, several macro packages, and a suite of C programs including a style sheet editor, an automatic page make-up system which replaces W ' s output mechanism, and a family of drivers. It solves many of the problems which make plain 'l$jX difficult to use for commercial journal and book publishing. This article explains in detail how Type & Set works. History of the project Type & Set has been under development at Informat Computer Communications since February 1987. Informat is the software development and typesetting division of Current Science (formerly Gower Academic Journals), a publishing house, and because the two companies share the same premises we have had constant access to users and their suggestions and criticism. Some ideas in Type & Set are tsken from an earlier (non-QX) package of the same name which it has superseded. For these ideas (principally the style sheet hierarchy, the mark-up system and the input format for the table generator) I am indebted to Mr. A. Harris, a former programmer at Informat. I take full responsibility for the present form of the system. The first version of Type & Set was installed in June 1988. but since then nearly every part has been rewritten. What problems does Type & Set solve? Using Type & Set rather than QX incurs costs in running time and disk space. However, Type & Set solves or ameliorates the following problems, many of which are discussed in detail by Mittelbach [I]. The severity of these problems amply justifies the increased use of resources. Page breaking is taken away from TEX completely and given to a program called PAGE which analyses the D V I file and writes a new D V I file. optimally paginated, with balanced columns, figure spaces, running material, headers and footers. PAGE takes i ts formatting information from a style sheet created using Type & Set's style sheet editor. Varying numbers of columns. Type & Set can 13 switch freely, as many times as you like (and as many times as you like on the same page) . between text in one. two, three and four columns. Baseline-to-baseline spacing occurs as a result of using PAGE rather than TEX to make up pages. All vertical dimensions in the Type & Set system are measured from the baseline of one line of text to the baseline of another. In particular. baselines at the bottom of pairs of columns align with each other, as do those of the last lines of text on facing pages. This also allows style sheets to specify a grid of lines on to which all baselines should fall if possible: that is, the y coordinate of a baseline should be an exact multiple of the grid interval. Composite fonts (I prefer this term, suggested by Beebe [ 2 ] . to the less descriptive 'virtuai fonts') are used where necessary in the drivers. Readable data files called FD or 'FontData' files provide all the information a driver needs to convert a QX character code into a device character, using transformations and superimposition if necessary. A utility. hfAKETFhl, is used to create TFM files for various output devices, given appropriate FD files and width tables. Tables are created using a quasi-wyszwyg format in a text editor and converted into TjjX by a program called TABLE. Horizontal spans, vertical and horizontal rules, and centring around any character (such as a decimal point), are all supported. Tables are very easily created and modified using this system. Graphics is absent from TEX, and should not be added. The prevailing standard for graphics is Postscript. and so the Type & Set Postscript driver will pick up a named Encapsulated Postscript file, translate and scale it, and embed it in a figure space. The driver is told to do this by a \special written by a macro placed in the text and passed through by PAGE. A more general feature of PAGE, that can be used with any driver, is its ability to load, scale and embed a D V I file in exactly the same way. Ease of use. Once a style sheet has been created for each kind of document to be typeset the rest is very easy. Staff at all levels of the publishing process, including those with no specialist computer knowledge (that is, nothing beyond basic abilities such as the use of the file system and rudimentary text editing) can be trained to use Type & Set in a day or two. The rest of this article describes in detail how the problems were solved and how Type & Set works. What is described is a working system which was designed and implemented at a publishing house over a period of four years, and is now in daily use. 14 TCGboat, Volume 13 (1992), No. 1 Data flow and general operation Input. The user types a document using his or her favourite word processor. This may. for example. be Wordstar or Wordperfect, or (as I prefer, being in part an unreconstructed TJ$ hacker) an ordinary ASCII text editor. The document contains little or no TEX apart from markup codes known as mode n a m e s which are determined by the style sheet to be used. Mode names look like ordinary T@X control sequences. mainly because that is what they are. The preferred Type & Set style places mode names on separate lines. The mode determines all the stylistic and structural parameters of the text: its font, justification, indents. paragraph spacing, and whether it is part of the body text. a figure caption, or, say. a running header-and many other details. Tag is the term preferred in the world of desktop publishing, but we stay with mode for historic reasons. Preprocessing. The first part of Type & Set to be run is the appropriate preprocessor for the text editor or word processor that has been used. In the case of Wordstar this is WS2TEX. which strips the high bits that Wordstar uses to mark the ends of words. converts Wordstar codes for italic, bold face, etc., into \it. \bf. etc., and emits standard ASCII text of the type 7&X reads. 'QX. Any version of can be used, with the proviso that if the document contains large tables a version with the biggest possible memory is desirable. loads a customised format file, very similar to p l a i n . fmt (indeed, almost upwardly compatible) called t s p l a i n . f mt. The first command T@X finds in its input file is something like \ input mystyle . s t y , which loads the style sheet, which defines all the mode names and other markup codes used in the document. TJ$ then runs normally and writes a DVI file, using the minimal output routine from t s p l a i n . fmt. This output file is effectively a galley in traditional typesetting terms, in that the text has been set in the desired fonts and counted or broken into lines, but has not yet been made up into pages. The D V I file contains numerous \specials, mostly for use by PAGE. Page make-up. PAGE reads the D V I file and analyses i t into lines. determining the mode of each line from a \special. A packet of information is built up for each line giving quick access to its mode, leading, position in the D V I file, and so on. At this point any automatzc materzal is added. This consists mainly of the spacing and rules which the style sheet specifies for insertion before or after certain modes. or between paragraphs, or around blocks of text. PAGE then uses the line information to write a new file, given the extension DVP, but precisely conforming to the D V I format, which contains the made-up pages. If TEX fonts were used the work of Type & Set proper could end here, and the DVP file could be typeset using a third-party driver. Previewing. Both D V I and DVP files for any printing device can be previewed on the screen using the Type & Set previewer, DVISCR. which draws characters using a device-independent vector font. This evades the problem of screen bitmap fonts not being available for, say, Optima on the Linotronic 100. Proofing. Proofing is generally done on a laser printer, using the Postscript or Hewlett Packard LaserJet driver as appropriate. All drivers have the ability to emulate a font not found on the output device by using a similar one that zs present: and the Postscript driver is especially optimised for emulation. Printing. The driver family includes drivers for Postscript, Hewlett Packard LaserJet, Linotype devices using CORA V, Chelgraph devices using ACE, and Agfa Compugraphic devices. All drivers share common code which reads the FD (FontData) file, interprets the D V I file, and implements the composite font system.