****************************************************************************** INSTALL - PXP, the XML parser for O'Caml ****************************************************************************** ============================================================================== The "pxp" package ============================================================================== ------------------------------------------------------------------------------ Prerequisites ------------------------------------------------------------------------------ PXP requires that the ocamlnet/netstring package [1] is already installed (version 0.98 required). PXP works only with O'Caml >= 3.07. The installation procedure defined in the Makefile requires findlib [2] to work [3]. PXP may be optionally compiled with support for Alain Frisch's patch of ocamllex called "wlex" [4]. There is now also support for ulex, Alain's Unicode-aware replacement for ocamllex (same link), which is simpler to build and now highly recommended. ------------------------------------------------------------------------------ Configuration ------------------------------------------------------------------------------ Beginning with PXP 1.1 it is necessary to configure the parser! Configuration is very simple, and in almost all cases it is sufficient to do ./configure in the top level directory of the distribution. It is possible to turn some options on or off using the -with-xxx and -without-xxx arguments. You can get a list by invoking ./configure -help: - -with-lex Enables the lexical analyzers ("lexers") generated by the ocamllex tool. You can specify which ocamllex-based lexers are created with the -lexlist option (see below). - -with-wlex Enables the lexical analyzer that works for UTF-8 as internal encoding, and that is based on Alain Frisch's wlex tool. It is relatively small and almost as fast as the ocamllex-based lexers. I recommend it if it is ok to install another library (wlex). - -with-wlex-compat Creates a compatibility package pxp-wlex that includes lexers for UTF8 and ISO-8859-1 (may be required to build old software) - -with-ulex Enables the lexical analyzer that works for UTF-8 as internal encoding, and that is based on Alain Frisch's ulex tool. It is relatively small, but a bit slower than the ocamllex-based lexers. ulex will supersede wlex soon. ulex is required for the preprocessor (see below). - -with-pp Enables the PXP preprocessor (installed as package pxp-pp). See the file PREPROCESSOR for details. The preprocessor requires ulex. - -lexlist <list-of-encodings> Specifies the character encodings to support by the ocamllex- based lexers. You need only the encodings you are going to use for the internal representation of the XML data in memory. It is not necessary to mention a character set here if you only want to read an external file. Note that utf8 is also provided by both -with-wlex and -with-ulex, and it is reasonable to omit it here if one of the mentioned options is in effect. Note that you need at least one lexical analyzer to use PXP as parser. ------------------------------------------------------------------------------ Compilation ------------------------------------------------------------------------------ The Makefile defines the following goals: - make all compiles with the bytecode compiler and creates various bytecode archives (suffix .cma) and some bytecode objects (suffix .cmo) in the directories below src. - make opt compiles with the native compiler and creates various native archives (suffixes .cmxa and .a) and some native objects (suffixes .cmx and .o) in the directories below src. ------------------------------------------------------------------------------ Installation ------------------------------------------------------------------------------ The Makefile defines the following goals: - make install installs the bytecode archives, the interface definitions, and if present, the native archives in the default location of findlib. Up to five packages may be installed: pxp, pxp-engine, pxp-lex-iso88591, pxp-lex-utf8, pxp-wlex. - make uninstall removes any of the mentioned packages Note: Previous versions of PXP had a compatibility API for the old "markup" distribution. This API is no longer supported. Upgrading to the PXP API is not very difficult. ------------------------------------------------------------------------------ Usage with the help of "findlib" ------------------------------------------------------------------------------ You can refer to the parser as the findlib package "pxp": ocamlfind ocamlc -package pxp ... Using "pxp" includes as much features of the parser as available, i.e. everything that has been configured. This may result in large executables. One possibility to reduce the code size is to specify Netstring-related predicates (e.g. netstring_only_iso); see the documentation of Netstring. Note that these predicates reduce the number of conversion tables for character encodings, and this means that not every character encoding of external files can be processed. Another way of reducing the size of executables is to link only selected parts of PXP. It is possible to specify which PXP subpackage is linked; for example, ocamlfind ocamlc -package pxp-engine,pxp-lex-iso88591 ... will only use "pxp-engine" (the core package) and the lexical analyzer "pxp-lex-iso88591", even if you have installed more PXP packages. ------------------------------------------------------------------------------ Linking with the archives directly ------------------------------------------------------------------------------ The following archives and objects may be used: - pxp_engine.cma: The core of PXP (always needed) - pxp_lex_iso88591.cma: The ocamllex-based lexical analyzer if you want to internally represent texts as ISO-8859-1. - pxp_lex_link_iso88591.cmo: Registers pxp_lex_iso88591 as lexer. - pxp_lex_utf8.cma: The ocamllex-based lexical analyzer if you want to internally represent texts as UTF-8. - pxp_lex_link_utf8.cmo: Registers pxp_lex_utf8 as lexer. - pxp_wlex.cma: The wlex-based lexical analyzer that works for both ISO-8859-1 and UTF-8 and results in smaller executables (but needs wlex). - pxp_wlex_link.cmo: Registers pxp_wlex as lexer. - pxp_top.cmo: Loading this module into the toploop installs several printers for PXP types. Note that you need at least one of the lexical analyzers if you want to parse XML texts. You do not need them if your program uses other features of PXP but not parsing. The archives containing the lexers are only linked into your executable if you also link the corresponding "register module". ============================================================================== The examples ============================================================================== In the "examples" directory you find several applications of PXP. They require that PXP has been installed using findlib. See the Makefiles in the directories for descriptions of "make" goals. ============================================================================== Trouble shooting ============================================================================== ------------------------------------------------------------------------------ Solaris ------------------------------------------------------------------------------ The "make" utility of Solaris does not work properly enough; there is a bug in it that prevents the so-called suffix rules from being recognized. There are two solutions: - Install GNU make and use it instead of Solaris make. This is the recommended way to solve the problem, as GNU make can process almost every Makefile from open source projects, and you will never have problems with building software again. - Add the following lines to Makefile.rules: %.cmx: %.ml $(OCAMLOPT) -c $< %.cmo: %.ml $(OCAMLC) -c $< %.cmi: %.mli $(OCAMLC) -c $< %.ml: %.mll ocamllex $< -------------------------- [1] see http://projects.camlcity.org/projects/ocamlnet.html [2] see http://projects.camlcity.org/projects/findlib.html [3] Findlib is a package manager, see the file ABOUT-FINDLIB. [4] see http://www.eleves.ens.fr:8080/home/frisch/soft