[mod] outline fails
This commit is contained in:
parent
dda4630730
commit
d0e746a45a
@ -1,2 +1,2 @@
|
||||
# pdfbox
|
||||
BaseX (10+) interface to Pdfbox 3
|
||||
# Pdfbox
|
||||
BaseX (10+) interface to [Pdfbox](https://pdfbox.apache.org/) 3
|
||||
|
||||
159
xquery/lib/pdfbox-lib.xqm
Normal file
159
xquery/lib/pdfbox-lib.xqm
Normal file
@ -0,0 +1,159 @@
|
||||
xquery version '3.1';
|
||||
(:~ pdfbox 3.0 https://pdfbox.apache.org/ interface library,
|
||||
requires pdfbox jar on classpath
|
||||
3.02 required tested with pdfbox-app-3.0.2-20240121.184204-66.jar
|
||||
@see https://lists.apache.org/list?users@pdfbox.apache.org:lte=1M:loader
|
||||
:)
|
||||
module namespace pdfbox="urn:expkg-zone58:pdfbox:3";
|
||||
|
||||
declare namespace Loader ="java:org.apache.pdfbox.Loader";
|
||||
declare namespace PDFTextStripper = "java:org.apache.pdfbox.text.PDFTextStripper";
|
||||
|
||||
(:~
|
||||
@see https://javadoc.io/static/org.apache.pdfbox/pdfbox/3.0.0/org/apache/pdfbox/pdmodel/PDDocument.html
|
||||
:)
|
||||
declare namespace PDDocument ="java:org.apache.pdfbox.pdmodel.PDDocument";
|
||||
|
||||
declare namespace PDDocumentCatalog ="java:org.apache.pdfbox.pdmodel.PDDocumentCatalog";
|
||||
declare namespace PDPageLabels ="java:org.apache.pdfbox.pdmodel.common.PDPageLabels";
|
||||
|
||||
(:~
|
||||
@see https://javadoc.io/static/org.apache.pdfbox/pdfbox/3.0.0/org/apache/pdfbox/multipdf/PageExtractor.html
|
||||
:)
|
||||
declare namespace PageExtractor ="java:org.apache.pdfbox.multipdf.PageExtractor";
|
||||
|
||||
(:~
|
||||
@see https://javadoc.io/static/org.apache.pdfbox/pdfbox/3.0.0/org/apache/pdfbox/pdmodel/PDPageTree.html
|
||||
:)
|
||||
declare namespace PDPageTree ="java:org.apache.pdfbox.pdmodel.PDPageTree";
|
||||
|
||||
(:~
|
||||
@see https://javadoc.io/static/org.apache.pdfbox/pdfbox/3.0.0/org/apache/pdfbox/pdmodel/interactive/documentnavigation/outline/PDDocumentOutline.html
|
||||
:)
|
||||
declare namespace PDDocumentOutline ="java:org.apache.pdfbox.pdmodel.interactive.documentnavigation.outline.PDDocumentOutline";
|
||||
|
||||
|
||||
(:~
|
||||
@see https://javadoc.io/static/org.apache.pdfbox/pdfbox/3.0.0/org/apache/pdfbox/pdmodel/interactive/documentnavigation/outline/PDOutlineItem.html
|
||||
:)
|
||||
declare namespace PDOutlineItem="java:org.apache.pdfbox.pdmodel.interactive.documentnavigation.outline.PDOutlineItem";
|
||||
|
||||
declare namespace File ="java:java.io.File";
|
||||
declare namespace RandomAccessReadBufferedFile = "java:org.apache.pdfbox.io.RandomAccessReadBufferedFile";
|
||||
|
||||
declare function pdfbox:open($pdfpath as xs:string){
|
||||
Loader:loadPDF( RandomAccessReadBufferedFile:new($pdfpath))
|
||||
};
|
||||
|
||||
declare function pdfbox:save($doc,$savepath as xs:string)
|
||||
as xs:string{
|
||||
PDDocument:save($doc,File:new($savepath)),$savepath
|
||||
};
|
||||
|
||||
declare function pdfbox:close($doc){
|
||||
PDDocument:close($doc)
|
||||
};
|
||||
|
||||
declare function pdfbox:page-count($doc as item())
|
||||
as xs:integer{
|
||||
PDDocument:getNumberOfPages($doc)
|
||||
};
|
||||
|
||||
(:~ @TODO
|
||||
PDOutlineItem current = bookmark.getFirstChild();
|
||||
while (current != null) {
|
||||
PDPage currentPage = current.findDestinationPage(document);
|
||||
Integer pageNumber = document.getDocumentCatalog().getPages().indexOf(currentPage) + 1;
|
||||
System.out.println(current.getTitle() + "-------->" + pageNumber);
|
||||
getOutlines(document, current, indentation);
|
||||
current = current.getNextSibling();
|
||||
}
|
||||
:)
|
||||
declare function pdfbox:siblings($acc as item()*,$outlineItem ,$doc as item())
|
||||
{
|
||||
(# db:wrapjava all #) {
|
||||
if(empty($outlineItem))
|
||||
then $acc
|
||||
else let $next:= PDOutlineItem:getNextSibling($outlineItem)=>trace("next: ")
|
||||
return pdfbox:siblings($acc,pdfbox:bookmark($outlineItem ,$doc),
|
||||
$next
|
||||
)
|
||||
}
|
||||
};
|
||||
|
||||
declare function pdfbox:outline($doc as item())
|
||||
as item()*{
|
||||
let $bookmark:=(# db:wrapjava all #) {
|
||||
PDDocument:getDocumentCatalog($doc)
|
||||
=>PDDocumentCatalog:getDocumentOutline()
|
||||
=>PDOutlineItem:getFirstChild()=>trace("cur")
|
||||
}
|
||||
(: return hof:until(empty#1,pdfbox:outx(?,$doc),()) :)
|
||||
let $bk:=pdfbox:siblings((),$bookmark ,$doc)
|
||||
|
||||
return $bk
|
||||
};
|
||||
|
||||
declare function pdfbox:bookmark($bookmark as item(),$doc as item())
|
||||
as map(*){
|
||||
let $currentPage := PDOutlineItem:findDestinationPage($bookmark,$doc)
|
||||
(: return hof:until(empty#1,pdfbox:outx(?,$doc),()) :)
|
||||
return map{
|
||||
"index": pdfbox:pageIndex($currentPage,$doc),
|
||||
"title": PDOutlineItem:getTitle($bookmark)
|
||||
}
|
||||
};
|
||||
|
||||
declare function pdfbox:outx($page,$document){
|
||||
let $currentPage := PDOutlineItem:findDestinationPage($page,$document)
|
||||
let $pageNumber := pdfbox:pageIndex($currentPage,$document)
|
||||
return $pageNumber
|
||||
};
|
||||
|
||||
(: pageIndex of $page in $document :)
|
||||
declare function pdfbox:pageIndex(
|
||||
$page (: as java:org.apache.pdfbox.pdmodel.PDPage :),
|
||||
$document)
|
||||
{
|
||||
PDDocument:getDocumentCatalog($document)
|
||||
=>PDDocumentCatalog:getPages()
|
||||
=>PDPageTree:indexOf($page)
|
||||
};
|
||||
|
||||
|
||||
|
||||
(:~ new PDF doc from 1 based page range :)
|
||||
declare function pdfbox:extract($doc as item(),$target as xs:string,
|
||||
$start as xs:integer,$end as xs:integer)
|
||||
{
|
||||
let $a:=PageExtractor:new($doc, $start, $end) =>PageExtractor:extract()
|
||||
let $map:=pdfbox:save($a,$target)
|
||||
return pdfbox:close($a)
|
||||
};
|
||||
|
||||
|
||||
(:~ @TODO
|
||||
@see https://codereview.stackexchange.com/questions/286078/java-code-showing-page-labels-from-pdf-files
|
||||
:)
|
||||
declare function pdfbox:getPageLabels($doc as item())
|
||||
as item()*{
|
||||
PDDocument:getDocumentCatalog($doc)
|
||||
=>PDDocumentCatalog:getPageLabels()
|
||||
=>PDPageLabels:getLabelsByPageIndices()
|
||||
};
|
||||
(:~ @TODO
|
||||
@see https://codereview.stackexchange.com/questions/286078/java-code-showing-page-labels-from-pdf-files
|
||||
:)
|
||||
|
||||
(: text on $pageNo :)
|
||||
declare function pdfbox:getText($doc as item(), $pageNo as xs:integer)
|
||||
as xs:string{
|
||||
let $tStripper := (# db:wrapjava instance #) {
|
||||
PDFTextStripper:new()
|
||||
=> PDFTextStripper:setStartPage($pageNo)
|
||||
=> PDFTextStripper:setEndPage($pageNo)
|
||||
}
|
||||
return (# db:checkstrings #) {PDFTextStripper:getText($tStripper,$doc)}
|
||||
};
|
||||
|
||||
|
||||
25
xquery/scratch/pdfbox.xq
Normal file
25
xquery/scratch/pdfbox.xq
Normal file
@ -0,0 +1,25 @@
|
||||
(: PDFBOX experiments
|
||||
:)
|
||||
|
||||
import module namespace pdfbox="urn:expkg-zone58:pdfbox:3" at "../lib/pdfbox-lib.xqm";
|
||||
|
||||
|
||||
declare variable $samples:= map{
|
||||
"climate": "data\drop-01d\set\2-6-1\A5579C_1\271989---Book_File-Web_PDF_9798400627484_486728.pdf",
|
||||
"women": "data\drop-01d\set\2-6-1\A6229C_1\257334---Book_File-Web_PDF_9798216172628_486742.pdf",
|
||||
"genocide": "data\drop1-pdf\GR2967-TRD\272791---Book_File-Web_PDF_9798400640216_486366.pdf",
|
||||
"world": "data\drop-01c\gpg-book\2-6\A3506C-TRD\256186---Book_File-Web_PDF_9798216038955_486148.pdf",
|
||||
"dummy": "lib\abc-clio-dummy.pdf"
|
||||
};
|
||||
declare variable $base:= "C:\Users\mrwhe\git\bloomsbury\content-architecture\xquery\ABC-CLIO\data";
|
||||
(:~ resolve :)
|
||||
declare variable $PDF:= $samples?climate=>file:resolve-path($base);
|
||||
|
||||
(: women pdfs :)
|
||||
(: let $pdfdoc:="data\drop-01d\set\2-6-1\A6229C_1\outputs\9798216172628\2798216172625\pdfs\chunks-docbook.xml"
|
||||
=>file:resolve-path(file:base-dir())
|
||||
=>doc() :)
|
||||
|
||||
let $doc:=pdfbox:open($PDF)
|
||||
return pdfbox:outline($doc)
|
||||
(: return pdfbox:extract($doc,"c:\tmp\junk2.pdf",50,100) :)
|
||||
Loading…
Reference in New Issue
Block a user