hello
Ideally, ssax-core prelude would be pulled out of ssax-core so it is usable with other pieces of Oleg's code. * Other resources ** web-site.ss "The web-site collection is a library for generating static web content with SXML and SXSLT templates. It makes use of the SSAX library, which is now included as a standard PLT collection." http://planet.plt-scheme.org/docs/dherman/web-site.plt/2/0/doc.txt ** Papers *** XML, XPath, XSLT implementations as SXML, SXPath, and SXSLT http://okmij.org/ftp/papers/SXs-talk.pdf How S-exprs for XML fit in a legacy (document and programmer) world. Discussion of SXPath, SXML, SXSLT, interoperability. A good introduction. http://okmij.org/ftp/papers/SXs.pdf *** SXSLT vs XSLT; walkthrough of SXSLT http://okmij.org/ftp/papers/SXSLT-talk.pdf http://pair.com/lisovsky/STX ?? ** Oleg's SXML->HTML extended examples such as xml.scm and SXML-to-HTML-ext.scm. http://okmij.org/ftp/Scheme/xml.html#XML-authoring ** sxpath in sxml-tools Implementation of XPath in Scheme. It looks much easier to use this than to pick out elements from an SXML tree by hand. home page is http://okmij.org/ftp/Scheme/xml.html#SXPath Examples at http://www196.pair.com/lisovsky/query/sxpath/ especially: http://www196.pair.com/lisovsky/query/sxpath/tutorial/tutorial4.scm -> saved to ~/scheme/sxml/sxpath-tutorial.scm You must include a *TOP* element for sxpath to work, but it can be any symbol: '(*TOP* (elt "text")) and '(zzz (elt "text")) are both searchable with (sxpath '(elt)), but '(elt "text") and '((elt "text")) are not. -> Each symbol element eg. elt, translates to (select-kids (node-typeof? 'elt)) so this is expected behaviour. *** Note on tags with multiple text items, such as (url "a.html" "A document") Works for one item: ((sxpath '(url *text*)) '(*TOP* (url "a.html" "A"))) ("a.html" "A") But results in flat list for more than one item: ((sxpath '(url *text*)) '(*TOP* (url "a.html" "A") (url "b.html" "B"))) ("a.html" "A" "b.html" "B") So don't request text element, request enclosing tag: ((sxpath '(url)) '(*TOP* (url "a.html" "A") (url "b.html" "B"))) ((url "a.html" "A") (url "b.html" "B")) Or, request an element number directly (this does 1st text element under url): ((sxpath '(url (*text* 1))) '(*TOP* (url "a.html" "A") (url "b.html" "B"))) ("a.html" "b.html") ((sxpath '(url (*text* 2))) '(*TOP* (url "a.html" "A") (url "b.html" "B"))) ("A" "B") Applying number to url itself gives you the entire second url tag, equivalent to (list (second ((sxpath '(url)) doc))): ((sxpath '((url 2))) '(*TOP* (url "a.html" "A") (url "b.html" "B"))) ((url "b.html" "B")) *** Chicken outdated version Oleg's code does not support textual XPath representation. -> Lisovsky et al have updated versions @ ssax.sf.net. -> That includes many other SXML related tools. Chicken's code is supposedly based off Lisovsky's version as of May 21 2004, but it only consists of "sxpath.scm" and is basically Oleg's code exactly. So I assume some screw-up occurred. (chicken post: http://lists.gnu.org/archive/html/chicken-users/2004-05/msg00076.html) But it never happened because on 7 Mar 2005 sxpath was posted again: http://lists.gnu.org/archive/html/chicken-users/2005-03/msg00009.html *** Porting ssax.sf.net version of sxpath. : Porting--starting from gambit version. util.scm -- there are a bunch of text files and scripts starting with util* I created, that analyse which functions in util.scm are used. Esp. utils-required.txt. Mainly string functions. ;; string-split : arg 2 is a list; chicken native is string. Can define string-split that automatically uses list->string, though this overrides a builtin. Requires a (no standard-bindings string-split) or whatever, and should probably be confined to that file only with a hide and/or a include. The problem boils down to, how to provide a function only within a group of source files and not externally to the user. Well, packages.scm has the goods. It separates everything into packages, exports what's necessary, pulls in only what's necessary (granularity is individual functions in a module). -> Believe this is Scheme48 syntax. -> References modules such as coutputs and assertions in http://download.plt-scheme.org/scheme/plt/collects/ssax/ -> Many of these in turn just source files from the ssax.sf.net distribution (in the same directory, plt/collects/ssax -> For example, coutputs.ss declares a mzscheme module and sources output.scm verbatim. Interesting technique. -- If I (require-extension coutputs) from sxpath.scm, when compiled the extension will not be loaded; instead it will be loaded at runtime, and made available to everyone. Cannot figure out how to load an extension only for one namespace. -- Note: basically everything from the scheme files extensions are exported everywhere, in packages.scm. The only things that should not be exported are coutputs and oleg-utils (for string-split and substring?) -- Therefore, these can freely be compiled separately, each exporting its entire interface. They should include coutputs and the string code, so that they are not exported sxpathlib srfi-13 ; string-prefix? string-index-right; only required for ntype-namespace-id?? ; assertions ; not required coutputs ; ppretty-print ; pp is native; display-circle is not used in the sxml-tools and need not be defined guide: not used anywhere (provide anyway?) assertions: -> Not required. assert is chicken native, and assure is not used xpath-ast.scm xpath-context.scm -> Not loaded by packages; are their functions referenced? oleg-string-ports: (provide with-output-to-string call-with-input-string with-input-from-string) -> which are already available natively in chicken. -> sxml-tools latest version replaced string-rindex with string-index-right from srfi-13. -> Must select "modularization" branch in CVS. Source changes: : Commented out define-syntax sxml:find-name-separator in sxml-tools.scm. : inexact->exact on round in sxpath-ext.scm (L173) **** Verification : csi -eval '(load "chicken/sxml-tools.scm") (load "chicken/tests-no-parent.scm") (load "tests/vsxpath-ext.scm")' ;; vsxpathlib.scm -- 1 test fails: (test-sxpath (quote (//))) (line 234) should be same as (sxml:descendant-or-self sxml:node?) test -- Apparently expected behavior. -- See http://sourceforge.net/mailarchive/message.php?msg_id=10210439 -- The code this expands to selects attribute nodes, while sxml:descendant-or-self is correct. -- Also see http://xmlhack.ru/protva/xquery/index.php/WhyNotSxTools -- Incorrect behavior demonstrated (as if it's correct) in example at http://www196.pair.com/lisovsky/query/examples/xpath/100.txt -- Seems on 2004/12/14 sxpath.scm was updated but my version doesn't work: http://sourceforge.net/mailarchive/message.php?msg_id=10300589 -- Must sync with MAIN branch, not 'modularization'. See below. ;; vtxpath.scm -- successful -- Test 1 was failing: sxml:test-xpointer+index"xpointer( */*[9]/*[5]/following::node() = //appendix//item[1] )" -> Error: (list-tail) bad argument type: 2.0 -> Fixed -- round in sxpath-ext.scm not using inexact->exact as in txpath.scm ;; vsxpath-ext.scm -- successful ;; vcontext.scm -- successful Requires xpath-context and xpath-ast. These use let-values* which is obsolete; added define-macro to beginning of sxml-tools.scm csi -eval '(use debug) (load "chicken/sxml-tools.scm") (load "chicken/tests-no-parent.scm") (load "xpath-ast.scm") (load "xpath-context.scm") (load "tests/vcontext.scm")' ;; vmodif.scm -- successful -- needs xpath-context and ddo. csi -eval '(use debug) (load "chicken/sxml-tools.scm") (load "chicken/tests-no-parent.scm") (load "xpath-ast.scm") (load "xpath-context.scm") (load "ddo-txpath.scm") (load "ddo-axes.scm") (load "modif.scm") (load "tests/vmodif.scm")' ;;; not implemented yet: ;; vddo.scm : csi -eval '(use debug) (load "chicken/sxml-tools.scm") (load "chicken/tests-no-parent.scm") (load "xpath-ast.scm") (load "xpath-context.scm") (load "ddo-txpath.scm") (load "ddo-axes.scm") (load "tests/vddo.scm")' Got unexpected result on (ddo:txpath doc/body/chapter[position()=2 or position()=5]/@id/ancestor-or-self::node()[position()<=3]/child::node()) ;; No tests for xlink, xlink-parser. **** Sync with latest CVS Latest development is being done on HEAD branch, not modularization, including fix for sxpath //. ddo-axes.scm: none ddo-txpath.scm: none lazy-xpath.scm: string-rindex modif.scm: none stx-engine.scm: sxml:error sxml-tools: remove sxml:error function; sxpath-ext.scm: let-values* sxpathlib.scm: string-rindex; uses (-- x) for (- x 1) sxpath-plus.scm: none sxpath.scm: none txpath.scm: sxml:error in sxml:xpointer-runtime-error (now exit -1); string-rindex; xlink-parser.scm: string-rindex xlink.scm: none xpath-ast.scm: none xpath-context.scm: none xpath-parser: used ascii->char; so ascii.scm probably no longer necessary Did not modify string-rindex, just included an alias macro. All tests now pass successfully. ** XML->SXML *** htmlprag html->shtml and shtml->html conversion, simpler than oleg's SXML->HTML. I wrote a wrapper around html->shtml which pretty prints the output--- by default it is barely human-editable. I formerly used write-shtml-as-html to create an HTML page from SHTML source in eggdoc, but now use my sxml-transforms egg. htmlprag can parse poorly-formed html, unlike SSAX:XML->SXML. It will retain all newlines and tabs, which may not be appropriate. (call-with-input-file "supportmail.html" (lambda (port) (html->shtml port))) *** Oleg's html-parser.scm Oleg's HTML parser does similar. Note that it is only an example and requires a bit of munging to be used as a library. -- http://okmij.org/ftp/Scheme/html-parser.scm This is not included in the ssax egg, but is available in the sxml-transforms egg under SSAX/examples/. However, it requires cout/cerr/nl (available in (sxml-tools egg under extras/output.scm) and a compatibility layer to match capitalization between the ssax egg (SSAX) and the example (ssax): ssax:read-char-data -> SSAX:read-char-data [in the ssax egg] ssax:complete-start-tag -- Note: ssax egg is v4.9 -- CVS version uses lowercase ssax. Hence the mismatch. This glue layer is in ~/scheme/ssax-html-parser.scm although it doesn't work. So htmlprag is the quick solution. *** SSAX:XML->SXML Gauche doc @ http://www.shiro.dreamhost.com/scheme/gauche/man/gauche-refe_304.html#SEC354. This parser is finicky. For example, parsing eggdoc's output "args.html", which is XHTML 1.0 Strict, complains: Error: (ssax-utils.scm, line 160) [SSAX: port args.html, at 354/84] [wf-entdeclared] broken for nbsp This is kind of discussed by http://www.w3.org/TR/REC-xml/#include-if-valid -- by default XML recognizes only amp, gt, lt, apos, and quot. These are defined in ssax:predefined-parsed-entities and checked in ssax:handle-parsed-entity of SSAX.scm. It seems is a better choice than If I try to define nbsp internally with ]> I get Warning: Internal DTD subset is not currently handled and the same problem with nbsp. See http://sourceforge.net/mailarchive/message.php?msg_id=10941360 for an explanation about not parsing DTDs. SSAX:XML->SXML is meant to perform parsing of -well-formed- XML documents, meaning those for which a doctype is not required. External doctypes are not parsed, and you can't declare an internal doctype (with permitted entities). And you can't pass in extra entities (perhaps you could override ssax:predefined-parsed-entities if you had to). Thus, you can't successfully parse . works fine, though. It's recommended you take the ssax:xml->sxml code and extend it to allow a passed in entities alist, if you want to get nbsp to work. *** Resume.xml -> sxml Converted resume.xml to sxml with ssax:xml->sxml since it was well-formed. (with-output-to-file "~/Desktop/resume.sxml" (lambda () (pp (call-with-input-file "~/Desktop/resume.xml" (lambda (port) (SSAX:XML->SXML port '() )))))) **** Extended sxpath example set using SXML resume #;1> (use sxml-tools) #;2> ((sxpath '(resume identity address city)) doc) ((city "Mount Prospect")) #;3> ((sxpath '(// address city)) doc) ((city "Mount Prospect")) #;4> ((sxpath "/resume/identity/address/city") doc) ((city "Mount Prospect")) #;5> ((sxpath '(// skillset skill name *text*)) doc) ("HP-UX" "Large sites" " High availability" "Performance analysis" "Perl and Python" "GUI design" "Open source" "C programming" "Documentation" "Web design") #;6> (apply print (intersperse #5 "\n")) HP-UX Large sites High availability Performance analysis Perl and Python GUI design Open source C programming Documentation Web design #;7> ((sxpath '(// abbr *text*)) doc) ("LVM" "NFS" "NIS" "SD-UX" "SSH" "HTML" "CSS" "DMZ" "LDAP" "NTP" "JDBC" "SQL") ** Sxml transformation *** SXSLT (Pre-post-order) Oleg's SXML-tree-trans.scm Transforms source SXML tree using transformation environment (stylesheet), which is an alist of nodes to transformers. "Post-order is equivalent to evaluating the SXML tree as if it were scheme code; pre-order is like Scheme macro expansion." **** sxslt-advanced.scm Test MAKE-TOC-ENTRIES: (make-toc-entries `((*section (1) "Section1" (p "This is section 1")) (*section (2) "Section 2" (*section (2 1) "Section2-1" (p "This is section 2-1"))))) => ((li "1" ". " "Section1" #f) (li "2" ". " "Section 2" (ul ((li "1.2" ". " "Section2-1" #f))))) ;; Notice extra parentheses and #f's; these will be discarded by SRV:send-reply. ;; You could discard extra parentheses with (pre-post-order-composable...