FOP supports multiple output formats by using a different renderer for each format. The renderers do not all have the same set of capabilities, sometimes because of the output format itself, sometimes because some renderers get more development attention than others.
Most FOP renderers use a FOP-specific system for font registration. However, the Java2D/AWT and print renderers use the Java AWT package, which gets its font information from the operating system registration. This can result in several differences, including actually using different fonts, and having different font metrics for the same font. The net effect is that the layout of a given FO document can be quite different between renderers that do not use the same font information.
Theoretically, there's some potential to make the output of the PDF/PS renderers match the output of the Java2D-based renderers. If FOP used the font metrics from its own font subsystem but still used Java2D for text painting in the Java2D-based renderers, this could probably be achieved. However, this approach hasn't been implemented, yet.
With a work-around, it is possible to match the PDF/PS output in a Java2D-based renderer pretty closely. The clue is to use the intermediate format. The trick is to layout the document using FOP's own font subsystem but then render the document using Java2D. Here are the necessary steps (using the command-line):
fop -fo myfile.fo -at application/pdf myfile.at.xmlfop -atin myfile.at.xml -pdf myfile.pdffop -atin myfile.at.xml -printfop -atin myfile.at.xml -awtfop -atin myfile.at.xml -tiff myfile.tiffThe most obvious way to print your document is to use the FOP print renderer, which uses the Java2D API (AWT). However, you can also send output from the Postscript renderer directly to a Postscript device, or output from the PCL renderer directly to a PCL device.
Here are Windows command-line examples for Postscript and PCL:
Here is some Java code to accomplish the task in UNIX:
Set the output MIME type to "application/x-pcl" (MimeConstants.MIME_PCL) and it happily sends the PCL to the UNIX printer queue.
PDF is the best supported output format. It is also the most accurate with text and layout. This creates a PDF document that is streamed out as each page is rendered. This means that the internal page index information is stored near the end of the document. The PDF version supported is 1.4. PDF versions are forwards/backwards compatible.
Note that FOP does not currently support "tagged PDF" or PDF/A-1a. Support for PDF/A-1b and PDF/X has recently been added, however.
PDF has a set of fonts that are always available to all PDF viewers; to quote from the PDF Specification: "PDF prescribes a set of 14 standard fonts that can be used without prior definition. These include four faces each of three Latin text typefaces (Courier, Helvetica, and Times), as well as two symbolic fonts (Symbol and ITC Zapf Dingbats). These fonts, or suitable substitute fonts with the same metrics, are guaranteed to be available in all PDF viewer applications."
FOP does not currently support several desirable PDF features: watermarks and signatures. One workaround is to use Adobe Acrobat (the full version, not the Reader) to process the file manually or with scripting that it supports.
Another popular post-processing tool is iText, which has tools for adding security features, document properties, watermarks, and many other features to PDF files.
Here is some sample code that uses iText to encrypt a FOP-generated PDF. (Note that FOP now supports PDF encryption. However the principles for using iText for other PDF features are similar.)
Check the iText tutorial and documentation for setting access flags, password, encryption strength and other parameters.
In addition to the PDF Post-processing options, consider the following workarounds:
The PostScript renderer has been brought up to a similar quality as the PDF renderer, but may still be missing certain features. It provides good support for most text and layout. Images and SVG are not fully supported, yet. Currently, the PostScript renderer generates PostScript Level 3 with most DSC comments. Actually, the only Level 3 features used are the FlateDecode and DCTDecode filter (the latter is used for 1:1 embedding of JPEG images), everything else is Level 2.
The PostScript renderer configuration currently allows the following settings:
The default value for the "auto-rotate-landscape" setting is "false". Setting it to "true" will automatically rotate landscape pages and will mark them as landscape.
The default value for the "language-level" setting is "3". This setting specifies the PostScript language level which should be used by FOP. Set this to "2" only if you don't have a Level 3 capable interpreter.
The default value for the "optimize-resources" setting is "false". Setting it to "true" will produce the PostScript file in two steps. A temporary file will be written first which will then be processed to add only the fonts which were really used and images are added to the stream only once as PostScript forms. This will reduce file size but can potentially increase the memory needed in the interpreter to process.
The default value for the "safe-set-page-device" setting is "false". Setting it to "true" will cause the renderer to invoke a postscript macro which guards against the possibility of invalid/unsupported postscript key/values being issued to the implementing postscript page device.
The default value for the "dsc-compliant" setting is "true". Setting it to "false" will break DSC compliance by minimizing the number of setpagedevice calls in the postscript document output. This feature may be useful when unwanted blank pages are experienced in your postscript output. This problem is caused by the particular postscript implementation issuing unwanted postscript subsystem initgraphics/erasepage calls on each setpagedevice call.
This format is for the Hewlett-Packard PCL printers and other printers supporting PCL. It should produce output as close to identical as possible to the printed output of the PDFRenderer within the limitations of the renderer, and output device.
The output created by the PCLRenderer is generic PCL 5, HP GL/2 and PJL. This should allow any device fully supporting PCL 5 to be able to print the output generated by the PCLRenderer. PJL is used to control the print job and switch to the PCL language. PCL 5 is used for text, raster graphics and rectangular fill graphics. HP GL/2 is used for more complex painting operations. Certain painting operations are done off-screen and rendered to PCL as bitmaps because of limitations in PCL 5.
The PCL renderer configuration currently allows the following settings:
The default value for the "rendering" setting is "speed" which causes borders to be painted as plain rectangles. In this mode, no special borders (dotted, dashed etc.) are available. If you want support for all border modes, set the value to "quality" as indicated above. This will cause the borders to be painted as bitmaps.
The default value for the "text-rendering" setting is "auto" which paints the base fonts using PCL fonts. Non-base fonts are painted as bitmaps through Java2D. If the mix of painting methods results in unwelcome output, you can set this to "bitmap" which causes all text to be rendered as bitmaps.
The default value for the "disable-pjl" setting is "false". This means that the PCL renderer usually generates PJL commands before and after the document in order to switch a printer into PCL language. PJL commands can be disabled if you set this value to "true".
You can control the output resolution for the PCL using the "target resolution" setting on the FOUserAgent. The actual value will be rounded up to the next supported PCL resolution. Currently, only 300 and 600 dpi are supported which should be enough for most use cases. Note that this setting directly affects the size of the output file and the print quality.
The PCL Renderer supports some PCL specific extensions which can be embedded into the input FO document. To use the extensions the appropriate namespace must be declared in the fo:root element like this:
The page-source extension attribute on fo:simple-page-master allows to select the paper tray the sheet for a particular simple-page-master is to be taken from. Example:
Note: the tray number is a positive integer and the value depends on the target printer. Not all PCL printers support the same paper trays. Usually, "1" is the default tray, "2" is the manual paper feed, "3" is the manual envelope feed, "4" is the "lower" tray and "7" is "auto-select". Consult the technical reference for your printer for all available values.
The output-bin extension attribute on fo:simple-page-master allows to
select the output bin into which the printed output should be fed. Example:
Note: the output bin number is a positive integer and the value depends on the target printer. Not all PCL printers support the same output bins. Usually, "1" is the upper output bin, "2" is the lower (rear) output bin. Consult the technical reference for your printer for all available values.
The duplex-mode extension attribute on fo:simple-page-master allows to select the duplex mode to be used for a particular simple-page-master. Example:
Note: the duplex is a positive integer and the value depends on the target printer. Not all PCL printers support duplexing. Usually, "0" is simplex, "1" is duplex (long-edge binding), "2" is duplex (short-edge binding). Consult the technical reference for your printer for all available values.
The FOP AFP Renderer deals with creating documents conforming to the IBM AFP document architecture also refered to as MO:DCA (Mixed Object Document Content Architecture).
The mapping of XSL-FO elements to the major MO:DCA structures is as follows:
| XSL-FO element | MO:DCA-P object |
|---|---|
| fo:root | Document |
| fo:page-sequence | Page Group |
| fo:simple-page-master | Page |
FOP creates exactly one Document per Printfile with an optional Resource Group at the beginning. FOP does not create document indices.
This list is most likely badly incomplete.
There are still a big number of older (or limited) MO:DCA/IPDS environments in production out there. AFP has grown in functionality over time and not every environment supports the latest features. We're trying to make AFP output work in as many environments as possible. However, to make AFP output work on older environments it is recommended to set to configuration to 1 bit per pixel (see below on how to do this). In this case, all images are converted to bi-level images using IOCA function set 10 (FS10) and are enclosed in page-segments since some implementation cannot deal with IOCA objects directly. If a higher number of bits per pixel is configured, FOP has to switch to at least FS11 which may not work everywhere.
The AFP Renderer requires special configuration particularly related to fonts. AFP Render configuration is done through the normal FOP configuration file. The MIME type for the AFP Renderer is application/x-afp which means the AFP Renderer section in the FOP configuration file looks like:
There are 4 font configuration variants supported:
A typical raster font configuration looks like:
An outline font configuration is simpler as the individual font size entries are not required. However, the characterset definition is now required within the afp-font element.
If "base-uri" is missing or a relative URI, the fonts are resolved relative to the font base URI specified in the configuration (or on the FopFactory).
A CID-keyed font (Type 0, double-byte outline font) configuration is much the same as an outline font. However, the characterset definition is now required within the afp-font element.
Note that the value of the encoding attribute in the example is the double-byte encoding 'UnicodeBigUnmarked' (UTF-16BE).
Experimentation has shown that the font metrics for the FOP built-in Base14 fonts are actually very similar to some of the IBM outline and raster fonts. In cases were the IBM font files are not available the base-uri attribute in the afp-font element can be replaced by a base14-font attribute giving the name of the matching Base14 font. In this case the AFP Renderer will take the font metrics from the built-in font.
By default, all manually configured fonts are embedded, unless they are matched in the
referenced-fonts section of the configuration file.
However, the default fonts shown above will not be embedded.
By default the AFP Renderer creates output with a resolution of 240 dpi. This can be overridden by the <renderer-resolution/> configuration element. Example:
By default the AFP Renderer converts all images to 8 bit grey level. This can be overridden by the <images/> configuration element. Example:
This will put images as RGB images into the AFP output stream. The default setting is:
Only the values "color" and "b+w" are allowed for the mode attribute.
The bits-per-pixel attribute is ignored if mode is "color". For "b+w" mode is must be 1, 4, or 8.
When the native attribute is specified and set to "true", all image resources will be natively injected into the datastream using an object container rather than being converted into an IOCA FS45 image. Support for native image formats (e.g. JPEG, TIFF, GIF) is not always available on printer implementations so by default this configuration option is set to "false".
Setting cmyk="true" on the images element will enable CMYK
colors. This will only have an effect if the color mode is set to "color". Example:
When the color mode is set to 1 bit (bi-level), the "dithering-quality" attribute can be used to select the level of quality to use when converting images to bi-level images. Valid values for this attribute are floating point numbers from 0.0 (fastest) to 1.0 (best), or special values: "minimum" (=0.0), "maximum" (1.0), "medium" (0.5, the default). For the higher settings to work as expected, JAI needs to be present in the classpath. If JAI is present, 0.0 results in a minimal darkness-level switching between white and black. 0.5 does bayer-based dithering and 1.0 will use error-diffusion dithering. The higher the value, the higher the quality and the slower the processing of the images.
By default, filled rectangles are painted using their given color using a PTOCA I-axis rule (DIR). But not all environments handle these colors correctly. That's why a setting is supported that paints the rectangles using an ordered dither pattern (bi-level) with an inline IOCA FS10 image that is used together with the "replicate and trim" mapping. The optional "shading" element can be used to control the shading mode. Its default value is "color". To enable the dithered mode, use "dithered". Example:
By default the AFP Renderer will place all data resource objects such as images within the document of the main output datastream. An external resource group file where document resources may be specified with the <resource-group-file/> configuration element. Example:
By default, bitmap image objects (or page segments derived from them) are put in the print-file-level resource group and GOCA graphics are inlined for compatibility with the AFP Workbench tool.
It is possible to override these defaults, either per image (see the afp:resource-level extension attribute below) or by specifying different defaults in the configuration:
"goca" refers to GOCA graphics and "bitmap" refers to IOCA images. The possible values for the attributes are "inline" and "print-file". In the future, additional possibilities may be added.
The AFP Renderer supports some AFP specific extensions which can be embedded into the input fo document. To use the extensions the appropriate namespace must be declared in the fo:root element like this:
The include-page-overlay extension element allows to define on a per simple-page-master basis a page overlay resource. Example:
The mandatory name attribute must refer to an 8 character (space padded) resource name that must be known in the AFP processing environment. Optional x and y attributes can be specified to place the Overlay at an offset from the top left of the page.
The include-page-segment extension element allows to define resource substitution for fo:external-graphics elements. Example:
The include-page-segment extension element can only occur within a simple-page-master. Multiple include-page-segment extension elements within a simple-page-master are allowed. The mandatory name attribute must refer to an 8 character (space padded) resource name that must be known in the AFP processing environment. The value of the mandatory src attribute is compared against the value of the src attribute in fo:external-graphic elements and if it is identical (string matching is used) in the generated AFP the external graphic is replaced by a reference to the given resource.
The effect here is that whenever FOP encounters the URI specified in the extension, it will effectively generate code to include the page segment with the given name instead of embedding the image referenced by the URI. The URI is still required as the underlying image serves as a provider for the intrinsic size of the image (At the moment, FOP is unable to extract the intrinsic size of the page segment from an AFP resource file). For the image to appear in an AFP viewer or to be printed, the AFP resource must be available on the target device. FOP does not embed the page segment in the generated file. Please also note that page segments cannot be scaled. They are always rendered in their intrinsic size.
The tag-logical-element extension element allows to injects TLEs into the AFP output stream. Example:
The tag-logical-element extension element can appear within a simple-page-master (page level) or it can appear as child of page-sequence (page group level). Multiple tag-logical-element extension elements within a simple-page-master or page-sequence are allowed. The name and value attributes are mandatory.
The no-operation extension provides the ability to carry up to 32K of comments or any other type of unarchitected data into the AFP output stream. Example:
The no-operation extension element can only occur within a simple-page-master. Multiple no-operation extension elements within a simple-page-master are allowed. The name attribute is mandatory.
The invoke-medium-map extension allows to generate IMM fields (Invoke Medium Map) in the generated AFP output. Example:
The invoke-medium-map element is allowed as child of fo:page-sequence (page group level) or fo:simple-page-master. It is NOT supported on document level (fo:root), yet. FOP also doesn't support specifying medium maps inside XML (using BMM/EMM). It can only reference an existing medium map by name. The medium map has to be constructed through different means and available on the target platform.
Apache FOP supports embedding an external form map resource in the
generated AFP output. This is done using the afp:include-form-map
extension. An example:
The afp:include-form-map is to be placed as a direct child of
fo:declarations. The name is an AFP resource name
(max. 8 characters) and the src attribute is the URI identifying the
external form map resource. When such a form map is embedded, you can use the
afp:invoke-medium-map extension (described above) to invoke any medium
map included in the form map.
The resource foreign attributes provides the ability to name and control where data object resources (e.g. images/scalable vector graphics) will reside in the AFP output. The afp foreign attributes are only used in conjuntion with <fo:external-graphic/> and <instream-foreign-object/>. Example:
The resource-level attribute where the resource object will reside in the AFP output datastream. The possible values for this are "inline", "print-file" and "external". When "external" is used a resource-group-file attribute must also be specified. Please refer to the Resource Level Defaults above to see what is used if the resource-level attribute is not specified.
JFOR, an open source XSL-FO to RTF converter has been integrated into Apache FOP. This will create an RTF (rich text format) document that will attempt to contain as much information from the XSL-FO document as possible. It should be noted that is not possible (due to RTF's limitations) to map all XSL-FO features to RTF. For complex documents, the RTF output will never reach the feature level from PDF, for example. Thus, using RTF output is only recommended for simple documents such as letters.
The RTF output follows Microsoft's RTF specifications and produces best results on Microsoft Word.
These are some known restrictions compared to other supported output formats (not a complete list):
This is primarily for testing and verification. The XML created is simply a representation of the internal area tree put into XML. We use that to verify the functionality of FOP's layout engine.
The other use case of the Area Tree XML is as FOP's "intermediate format". More information on that can be found on the page dedicated to the Intermediate Format.
The Java2DRenderer provides the basic functionality for all Java2D-based output formats (AWT viewer, direct print, PNG, TIFF).
The AWT viewer shows a window with the pages displayed inside a Java graphic. It displays one page at a time. The fonts used for the formatting and viewing depend on the fonts available to your JRE.
It is possible to directly print the document from the command line. This is done with the same code that renders to the Java2D/AWT renderer.
If you run into the problem that the printed output is incomplete on Windows: this often happens to users printing to a PCL printer. There seems to be an incompatibility between Java and certain PCL printer drivers on Windows. Since most network-enabled laser printers support PostScript, try switching to the PostScript printer driver for that printer model.
It is possible to directly create bitmap images from the individual pages generated by the layout engine. This is done with the same code that renders to the Java2D/AWT renderer.
Currently, two output formats are supported: PNG and TIFF. TIFF produces
one file with multiple pages, while PNG output produces one file per
page. Note: FOP can only produce multiple files (with PNG output) if
you can set a java.io.File indicating the primary PNG file
using the FOUserAgent.setOutputFile(File) method.
The quality of the bitmap depends on the target resolution setting on the FOUserAgent and on further settings described below.
The TIFF and PNG renderer configuration currently allows the following settings:
The default value for the "color-mode" setting is "rgba" which
is equivalent to a 24bit RGB image with an 8bit alpha channel for transparency.
Valid values are:
rgba: RGB with alpha channel (24bit + 8bit = 32bit)rgb: RGB (24bit)gray: gray (8bit)bi-level (or binary): bi-level (1bit)Please note that there is currently no dithering or error diffusion available for bi-level bitmap output.
The default value for the "transparent-page-background" setting is
"false" which paints an opaque, white background for the whole image.
If you set this to "true",
no such background will be painted and you will get a transparent image if
an alpha channel is available in the output format.
The default value for the "background-color" setting is "white".
The color specifies in which color the page background is painted. It will only be
painted if "transparent-page-background" is not set to "true".
All XSL-FO colors (including color functions) can be used.
The default value for the "anti-aliasing" setting is "true".
You can set this value to "false" to disable anti-aliasing and
thus improve rendering speeds a bit at the loss of some image quality.
The default value for the "rendering" setting is "true".
You can set this value to "false" to improve rendering speeds a bit
at the loss of some image quality. If this setting has an actual effect depends
on the JVM's Java2D backend.
In addition to the above values the TIFF renderer configuration allows some additional settings:
The default value for the "compression" setting is "PackBits" which which is a widely supported RLE compression scheme for TIFF. The set of compression names to be used here matches the set that the Image I/O API uses. Note that not all compression schemes may be available during runtime. This depends on the actual codecs being available. Here is a list of possible values:
NONE (no compression)PackBits (RLE, run-length encoding)JPEGDeflateLZWZLibCCITT T.4 (Fax Group 3)CCITT T.6 (Fax Group 4)
This setting may override any setting made using the "color-mode". For example, if
"CCITT T.6" is selected, the color mode is automatically forced to "bi-level" because
this compression format only supports bi-level images.
The IF-based bitmap output implementations support a rendering option with the key "target-bitmap-size" (value: java.awt.Dimension) that allows to force the pages to be proportionally fit into a bitmap of a given size. This can be used to produce thumbnails or little preview images of the individual pages. An example:
The text renderer produces plain ASCII text output that attempts to match the output of the PDFRenderer as closely as possible. This was originally developed to accommodate an archive system that could only accept plain text files, and is primarily useful for getting a quick-and-dirty view of the document text. The renderer is very limited, so do not be surprised if it gives unsatisfactory results.
The Text renderer works with a fixed size page buffer. The size of this buffer is controlled with the textCPI and textLPI public variables. The textCPI is the effective horizontal characters per inch to use. The textLPI is the vertical lines per inch to use. From these values and the page width and height the size of the buffer is calculated. The formatting objects to be rendered are then mapped to this grid. Graphic elements (lines, borders, etc) are assigned a lower priority than text, so text will overwrite any graphic element representations.
Because FOP lays the text onto a grid during layout, there are frequently extra or missing spaces between characters and lines, which is generally unsatisfactory. Users have reported that the optimal settings to avoid such spacing problems are:
Due to the state of certain renderers we moved some of them to a "sandbox" area until they are ready for more serious use. The renderers and FOEventHandlers in the sandbox can be found under src/sandbox and are compiled into build/fop-sandbox.jar during the main build. The output formats in the sandbox are marked as such below.
This format is the Maker Interchange Format which is used by Adobe Framemaker.
This format creates an SVG document that has links between the pages. This is primarily for slides and creating svg images of pages. Large documents will create SVG files that are far too large for an SVG viewer to handle. Since FO documents usually have text the SVG document will have a large number of text elements. The font information for the text is obtained from the JVM in the same way as for the AWT viewer. If the SVG is viewed on a system where the fonts are different, such as another platform, then the page may look wrong.
Apache FOP is easily extensible and allows you to add new output formats to enhance FOP's functionality. There's a number of output formats which are on our wish list. We're looking for volunteers to help us implement them.