From a0cfa6d937d74f48e032fa1cfc15c366decf6c00 Mon Sep 17 00:00:00 2001
From: Andy Bunce <bunce.andy@gmail.com>
Date: Thu, 30 Jan 2025 16:57:25 +0000
Subject: [PATCH] [mod] download

---
 scripts/build.xqm       | 22 ++++++++--
 scripts/download.xq     | 17 --------
 scripts/make-fat-jar.xq |  7 +++
 src/Pdfbox3.xqm         | 96 +++++++++++++++++++++++------------------
 src/test/test.xqm       | 48 +++++++++++++--------
 5 files changed, 110 insertions(+), 80 deletions(-)
 delete mode 100644 scripts/download.xq

diff --git a/scripts/build.xqm b/scripts/build.xqm
index 9cfa38b..4e53ba3 100644
--- a/scripts/build.xqm
+++ b/scripts/build.xqm
@@ -1,7 +1,9 @@
 (:~ build utils for REPO packaging :)
 module namespace build = 'urn:quodatum:build1';
 
-(:~ create a flat fat jar :)
+(:~ create a flat fat jar from jars in $input-dir
+keeping only META-INF from $manifest-jar 
+:)
 declare function build:fatjar-from-folder($input-dir as xs:string,$manifest-jar as xs:string)
 as xs:base64Binary { 
     let $fold :=
@@ -13,13 +15,16 @@ function ($res as map (*), $jar as xs:string) {
         map { "name" : ($res? name, $paths), 
               "content" : ($res? content,archive:extract-binary($bin, $paths)) } 
 }
-let $res := fold-left(file:list($input-dir, false(), "*.jar"), map { }, $fold)
+let $res := file:list($input-dir, false(), "*.jar")
+            =>fold-left( map { }, $fold)
 return
     archive:create($res? name, $res? content,
                    map { "format" : "zip", "algorithm" : "deflate" }) 
 };
 
-(:~ create a fat jar with lib :)
+(:~ create a fat jar with lib 
+@remark 
+:)
 declare function build:fatjar-with-lib($input-dir as xs:string,$manifest-jar as xs:string)
  { 
  let $bin :=file:read-binary($input-dir || $manifest-jar)
@@ -49,3 +54,14 @@ declare function build:update($jar as xs:base64Binary,$name  as xs:string,$file
 as xs:base64Binary{
 archive:update($jar,$name,$file)
 }; 
+
+
+(:~ download $files from $urls to  $destdir:)
+declare variable $build:REPO as xs:string external :="https://repo1.maven.org/maven2/";
+declare function build:maven-download($urls as xs:string*,$destdir as xs:string)
+as empty-sequence(){
+for $f in $urls
+let $dest:=$destdir || replace($f,"^.*/","") 
+where not(file:exists($dest))
+return file:write-binary($dest, fetch:binary(resolve-uri($f,$build:REPO)=>trace("Download: ")))
+};
\ No newline at end of file
diff --git a/scripts/download.xq b/scripts/download.xq
deleted file mode 100644
index 3004e5f..0000000
--- a/scripts/download.xq
+++ /dev/null
@@ -1,17 +0,0 @@
-
-import module namespace build = 'urn:quodatum:build1' at 'build.xqm';
-
-declare variable $files := (
-"https://repo1.maven.org/maven2/org/apache/pdfbox/pdfbox/3.0.4/pdfbox-3.0.4.jar",
-"https://repo1.maven.org/maven2/org/apache/pdfbox/pdfbox-io/3.0.4/pdfbox-io-3.0.4.jar",
-"https://repo1.maven.org/maven2/org/apache/pdfbox/fontbox/3.0.4/fontbox-3.0.4.jar",
-"https://repo1.maven.org/maven2/commons-logging/commons-logging/1.3.4/commons-logging-1.3.4.jar"
-);
-
-
-
-let $base:= file:resolve-path("../",static-base-uri())
-let $target:=file:resolve-path("jars/",$base)
-for $f in $files
-let $n:=replace($f,"^.*/","") =>trace("N")
-return file:write-binary($target || $n, fetch:binary($f))
\ No newline at end of file
diff --git a/scripts/make-fat-jar.xq b/scripts/make-fat-jar.xq
index 51b8b16..a1cb737 100644
--- a/scripts/make-fat-jar.xq
+++ b/scripts/make-fat-jar.xq
@@ -1,6 +1,12 @@
 
 import module namespace build = 'urn:quodatum:build1' at 'build.xqm';
 
+declare variable $urls := (
+"org/apache/pdfbox/pdfbox/3.0.4/pdfbox-3.0.4.jar",
+"org/apache/pdfbox/pdfbox-io/3.0.4/pdfbox-io-3.0.4.jar",
+"org/apache/pdfbox/fontbox/3.0.4/fontbox-3.0.4.jar",
+"commons-logging/commons-logging/1.3.4/commons-logging-1.3.4.jar"
+);
 (: Main execution
 Main-Class: org.basex.modules.Hello
  :)
@@ -12,6 +18,7 @@ let $config :=map {
          "main-class": "org.expkg_zone58.Pdfbox3" 
          }
 let $jar-path:=file:resolve-path($config?input-dir,$config?base=>trace("base "))=>trace("jar: ")
+let $_:=build:maven-download($urls,$jar-path)
 let $fat-jar := build:fatjar-from-folder($jar-path,$config?manifest-jar)
 
 let $fat-jar:=build:update-manifest($fat-jar, $config?main-class)
diff --git a/src/Pdfbox3.xqm b/src/Pdfbox3.xqm
index a14a01a..a2a9315 100644
--- a/src/Pdfbox3.xqm
+++ b/src/Pdfbox3.xqm
@@ -49,10 +49,12 @@ as item(){
   Loader:loadPDF( RandomAccessReadBufferedFile:new($pdfpath))
 };
 
-(:~ the version of the PDF specification used by $pdf :)
-declare function pdfbox:pdfVersion($pdf as item())
-as xs:float{
-  PDDocument:getVersion($pdf)
+(:~ the version of the PDF specification used by $pdf  e.g "1.4"
+returned as string to avoid rounding issues
+ :)
+declare function pdfbox:specification($pdf as item())
+as xs:string{
+ PDDocument:getVersion($pdf)=>string()
 };
 
 (:~ save pdf $pdf to $savepath , returns $savepath :)
@@ -97,45 +99,49 @@ as xs:string{
  Q{java:java.util.GregorianCalendar}toZonedDateTime($item)=>string()
 };
 
-(:~ outline for $doc as map()* :)
-declare function pdfbox:outline($doc as item())
+(:~ outline for $pdf as map()* :)
+declare function pdfbox:outline($pdf as item())
 as map(*)*{
   (# db:wrapjava some #) {
   let $outline:=
-                PDDocument:getDocumentCatalog($doc)
+                PDDocument:getDocumentCatalog($pdf)
                 =>PDDocumentCatalog:getDocumentOutline()
  
   return  if(exists($outline))
-          then pdfbox:outline($doc,PDOutlineItem:getFirstChild($outline)) 
+          then pdfbox:outline($pdf,PDOutlineItem:getFirstChild($outline)) 
   }
 };
 
 (:~ return bookmark info for children of $outlineItem as seq of maps :)
-declare function pdfbox:outline($doc as item(),$outlineItem as item()?)
+declare function pdfbox:outline($pdf as item(),$outlineItem as item()?)
 
 as map(*)*{
-  let $find as map(*):=pdfbox:_outline($doc ,$outlineItem)
+  let $find as map(*):=pdfbox:_outline($pdf ,$outlineItem)
   return map:get($find,"list")
 };
 
 (: BaseX bug 10.7? error if inlined in outline :)
-declare %private function pdfbox:_outline($doc as item(),$outlineItem as item()?)
+declare %private function pdfbox:_outline($pdf as item(),$outlineItem as item()?)
 as map(*){
- hof:until(
-            function($output) { empty($output?this) },
-            function($input ) { 
-                      let $bk:= pdfbox:bookmark($input?this,$doc)
+  pdfbox:do-until(
+    
+     map{"list":(),"this":$outlineItem},
+
+     function($input ) { 
+                      let $bk:= pdfbox:bookmark($input?this,$pdf)
                       let $bk:= if($bk?hasChildren)
-                                then let $kids:=pdfbox:outline($doc,PDOutlineItem:getFirstChild($input?this))
+                                then let $kids:=pdfbox:outline($pdf,PDOutlineItem:getFirstChild($input?this))
                                      return map:merge(($bk,map:entry("children",$kids)))
                                 else $bk 
                       return map{
                             "list": ($input?list, $bk),
                             "this":  PDOutlineItem:getNextSibling($input?this)}
                           },
-            map{"list":(),"this":$outlineItem}
-        ) 
+
+     function($output) { empty($output?this) }                      
+  )
 };
+
 (:~ outline as xml :)
 declare function pdfbox:outline-xml($outline as map(*)*)
 as element(outline){
@@ -156,27 +162,22 @@ as element(bookmark)*
 (:~ return bookmark info for children of $outlineItem 
 @return map like{index:,title:,hasChildren:}
 :)
-declare function pdfbox:bookmark($bookmark as item(),$doc as item())
+declare function pdfbox:bookmark($bookmark as item(),$pdf as item())
 as map(*)
 {
  map{ 
-  "index":  PDOutlineItem:findDestinationPage($bookmark,$doc)=>pdfbox:pageIndex($doc),
+  "index":  PDOutlineItem:findDestinationPage($bookmark,$pdf)=>pdfbox:page-index($pdf),
   "title":  (# db:checkstrings #) {PDOutlineItem:getTitle($bookmark)}=>translate("�",""),
   "hasChildren": PDOutlineItem:hasChildren($bookmark)
   }
 };
 
-declare function pdfbox:outx($page ,$document)
-{
-  let $currentPage := PDOutlineItem:findDestinationPage($page,$document)
-  let $pageNumber := pdfbox:pageIndex($currentPage,$document)
-  return $pageNumber
-};
+
 
 (:~ pageIndex of $page in $pdf :)
-declare function pdfbox:pageIndex(
+declare function pdfbox:page-index(
    $page as item()? (: as java:org.apache.pdfbox.pdmodel.PDPage :),
-   $pdf)
+   $pdf as item())
 as item()?
 {
   if(exists($page))
@@ -198,28 +199,20 @@ as xs:string
 };
 
 
-(:~   pageLabel info
+(:~   pageLabel for every page
 @see https://www.w3.org/TR/WCAG20-TECHS/PDF17.html#PDF17-examples
 @see https://codereview.stackexchange.com/questions/286078/java-code-showing-page-labels-from-pdf-files
 :)
-declare function pdfbox:getPageLabels($pdf as item())
-as item()
-{
-  PDDocument:getDocumentCatalog($pdf)
-  =>PDDocumentCatalog:getPageLabels()
-};
-
-(:~   pageLabel for every page:)
-declare function pdfbox:pageLabels($doc as item())
+declare function pdfbox:labels($pdf as item())
 as xs:string*
 {
-  PDDocument:getDocumentCatalog($doc)
+  PDDocument:getDocumentCatalog($pdf)
   =>PDDocumentCatalog:getPageLabels()
   =>PDPageLabels:getLabelsByPageIndices()
 };
 
 (:~ return text on $pageNo :)
-declare function pdfbox:getText($doc as item(), $pageNo as xs:integer)
+declare function pdfbox:page-text($doc as item(), $pageNo as xs:integer)
 as xs:string{
   let $tStripper := (# db:wrapjava instance #) {
          PDFTextStripper:new()
@@ -246,9 +239,9 @@ as map(*){
 @param $scale 1=72 dpi 
 @return  Java java.awt.image.BufferedImage object
 :)
-declare function pdfbox:pageBufferedImage($doc as item(), $pageNo as xs:integer,$scale as xs:float)
+declare function pdfbox:pageBufferedImage($pdf as item(), $pageNo as xs:integer,$scale as xs:float)
 as item(){
- PDFRenderer:new($doc)=>PDFRenderer:renderImage($pageNo,$scale)
+ PDFRenderer:new($pdf)=>PDFRenderer:renderImage($pageNo,$scale)
 };
 
 (:~ save bufferedimage to $dest 
@@ -266,4 +259,21 @@ as xs:base64Binary{
   let $_:=Q{java:javax.imageio.ImageIO}write($bufferedImage , $type,  $bytes)
   return Q{java:java.io.ByteArrayOutputStream}toByteArray($bytes)
          =>convert:integers-to-base64()
-};
\ No newline at end of file
+};
+
+(:~ fn:do-until shim for BaseX 9+ :)
+declare function pdfbox:do-until(
+ $input 	as item()*, 	
+ $action 	as function(item()*, xs:integer) as item()*, 	
+ $predicate 	as function(item()*, xs:integer) as xs:boolean? 	
+) as item()*
+{
+  let $fn:=function-lookup(QName('http://www.w3.org/2005/xpath-functions','do-until'), 3)
+  return if($fn)
+         then $fn($input,$action,$predicate)
+         else let $hof:=function-lookup(QName('http://basex.org/modules/hof','until'), 3)
+              return if($hof)
+                      then $hof($predicate,$action,$input)
+                      else error(xs:QName('pdfbox:do-until'),"No implementation found")
+
+};
diff --git a/src/test/test.xqm b/src/test/test.xqm
index bfd1f71..7b0958b 100644
--- a/src/test/test.xqm
+++ b/src/test/test.xqm
@@ -7,39 +7,52 @@ import module namespace pdfbox="org.expkg_zone58.Pdfbox3";
 declare variable $test:base:=file:base-dir()=>file:parent()=>file:parent();
 
 
+declare %unit:test
+function test:pdfbox-version(){
+    let $v:= pdfbox:version()=>trace("VER: ")
+    return unit:assert-equals($v,"3.0.4")
+};
+
+declare %unit:test
+function test:specification(){
+    let $pdf:=test:pdf("samples.pdf/BaseX100.pdf")
+    let $spec:=pdfbox:specification($pdf)
+    return unit:assert-equals($spec,0+1.4)
+};
 
 declare %unit:test
 function test:page-count(){
-    let $PDF:="samples.pdf/BaseX100.pdf"=>test:resolve()
-    let $pages:=pdfbox:open($PDF)=>pdfbox:page-count()
+    let $pdf:=test:pdf("samples.pdf/BaseX100.pdf")
+    let $pages:=pdfbox:page-count($pdf)
     return unit:assert-equals($pages,521)
 };
 
 declare %unit:test
 function test:outline-none(){
- let $PDF:="samples.pdf/BaseX100.pdf"=>test:resolve()
- let $outline:=pdfbox:open($PDF)=>pdfbox:outline()
+let $pdf:=test:pdf("samples.pdf/BaseX100.pdf")
+ let $outline:=pdfbox:outline($pdf)
  return unit:assert(empty($outline))
 };
 
 declare %unit:test
 function test:outline-present(){
- let $PDF:="samples.pdf/icelandic-dictionary.pdf"=>test:resolve()
- let $outline:=pdfbox:open($PDF)=>pdfbox:outline()
+ let $pdf:=test:pdf("samples.pdf/icelandic-dictionary.pdf")
+ let $outline:=pdfbox:outline($pdf)
  return unit:assert(exists($outline))
 };
 
 declare %unit:test
 function test:outline-xml(){
- let $PDF:="samples.pdf/icelandic-dictionary.pdf"=>test:resolve()
- let $outline:=pdfbox:open($PDF)=>pdfbox:outline()=>pdfbox:outline-xml()
+ let $pdf:=test:pdf("samples.pdf/icelandic-dictionary.pdf")
+ let $outline:=pdfbox:outline($pdf)=>pdfbox:outline-xml()
  return unit:assert-equals(count($outline/bookmark),31)
 };
 
 declare %unit:test
 function test:pagelabels(){
- let $PDF:="samples.pdf/BaseX100.pdf"=>test:resolve()
- let $labels:=pdfbox:open($PDF)=>pdfbox:pageLabels()
+  let $pdf:=test:pdf("samples.pdf/BaseX100.pdf")
+
+ let $labels:=pdfbox:labels($pdf)
  return (
    unit:assert($labels[1]="i") ,
    unit:assert($labels[27]="1")
@@ -47,20 +60,21 @@ function test:pagelabels(){
 };
 
 declare %unit:test
-function test:save(){
+function test:extract-save(){
+ let $pdf:=test:pdf("samples.pdf/BaseX100.pdf")
  let $dest:=file:create-temp-file("test",".pdf")=>trace("DEST: ")
- let $PDF:="samples.pdf/BaseX100.pdf"=>test:resolve()
- let $outline:=pdfbox:open($PDF)=>pdfbox:extract(2,12,$dest)
+ let $outline:=pdfbox:extract($pdf,2,12,$dest)
  return unit:assert(true())
 };
 
 declare %unit:test
 function test:page-text(){
- let $PDF:="samples.pdf/BaseX100.pdf"=>test:resolve()
- let $text:=pdfbox:open($PDF)=>pdfbox:getText(1)
+let $pdf:=test:pdf("samples.pdf/BaseX100.pdf")
+ let $text:=pdfbox:page-text($pdf,1)
  return unit:assert(starts-with($text,"BaseX Documentation"))
 };
 
-declare function test:resolve($file as xs:string){
-    file:resolve-path($file,$test:base)
+declare function test:pdf($file as xs:string)
+as item(){
+    file:resolve-path($file,$test:base)=>pdfbox:open()
 };