Introduction

Various pieces of software and techniques exist for using TEX to put labels onto included graphics. All have significant draw-backs or short-comings. One method that is widely used, and often recommended as best for Encapsulated POSTSCRIPT files, is to first Typeset the label using TEXTURES on a Macintosh, Copy the resulting typeset window, then Paste the clipboard contents into the image file, having been opened within Adobe's Illustrator application.

Among the draw-backs of this technique are:

• dependence upon a particular computing platform: Macintosh, or PowerMac;
• use of expensive commercial software: Adobe's Illustrator, and Blue Sky's TEXTURES;
• applicability to just a particular image format: Encapsulated POSTSCRIPT ;
• the original image file must be altered (after copying, please!) to obtain the required results.

Depending upon the working enviroment, these may not be problems at all; for example, a prepress house would be expected to have the appropriate hardware and software. Similarly an academic may have made the investment to be able to follow this strategy.

However there is a further problem which may cause great difficulties, when a manuscript is submitted for publication. Suppose a labelled image needs to be resized, or the labels need to be changed for some reason; e.g. the text-style chosen does not blend well with the fonts and styles used elsewhere in the publication. Now the Encapsulated POSTSCRIPT file needs to be edited, or regenerated in the same way as was done originally. This may no longer be possible--the software that created it may not be available or the expertise to use it may have been lost, due to the original operator being no longer around (having graduated or taken another job).

The WARMreader solution is to use TEX itself, or L^ATEX, for placing the labels. It uses the X_Y-pic diagram macros, extending the methods presented at the TUG'97 meeting, described in [3] and is available on the Web. Here the idea is to create a coordinate system specially tailored to the size of the imported image, then to anchor labels to appropriate places using these coordinates. This effectively creates an overlay which allows the labels to seem to be part of the image, when in actual fact the labels have been typeset by TEX. WARMreader takes this idea a step further, by automating the process to the extent that a user does not have to be concerned with the actual coordinates, when specifying the labels within the TEX or L^ATEX document.

These methods can be used on any computing platform. Since the styles and content of the labels are specified within the TEX or L^ATEX source, there is no need to alter any Encapsulated POSTSCRIPT files. Furthermore, these techniques can be used for graphics of any format that can be included within a TEX document, by whatever means. The only requirement is the ability to create a .bb file, containing information in an appropriate form.

For L^ATEX, the PSfrag package, as described in The L^ATEX Graphics Companion[1, pp.460-462], provides similar functionality when using Encapsulated POSTSCRIPT files, by treating parts of the POSTSCRIPT file as tags, to be later replaced by blocks of TEX-typeset material. The technique has several limitations, apart from being available only for L^ATEX, and not usable with graphics formats other than POSTSCRIPT. For best results the POSTSCRIPT file ``should ideally be designed with PSfrag in mind'', and for systematic use it ``requires a good understanding of both the POSTSCRIPT language and the application generating the figures''. This is because the replacement portions effectively become part of the POSTSCRIPT graphic at the point where the tags occur, so are subject to, and must dovetail with, the POSTSCRIPT graphics state at those places. As this includes color, size, rotation and cropping-region, great care is required to avoid later parts of the graphic obscuring earlier labels, or labels being cut-off at edges of the graphic. Indeed it is not possible to know exactly how the whole thing will appear until the DVI-file has been processed with a POSTSCRIPT-aware viewer or printer, thus making it very tedious to fine-tune the placement of labels.

With WARMreader, the labels can be regarded as occurring within a separate layer, controlled completely from within TEX or L^ATEX. Any graphic from any source, in any format that can be handled by the TEX installation, can be used as a ``backdrop'', provided that a suitable .bb file has been prepared. Each of the three steps:

1. construction of the graphic;
2. preparation of a .bb file, perhaps including text for labels;
3. preparation of code for processing labels within the TEX document;

can be done quite independently; e.g., by different people using different software or techniques. Only the last requires knowledge of TEX or L^ATEX, though this is desirable if labels are to be completely specified in the .bb file. Indeed it will become apparent below that the greatest control over the final appearance, hence the best results, are obtained when these three tasks are kept completely separate.