[
https://issues.apache.org/jira/browse/PDFBOX-1808?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13849213#comment-13849213
]
Tilman Hausherr commented on PDFBOX-1808:
-----------------------------------------
The nps file is the most interesting: look at the "live" columns only. Yes
you'll find much strings and chars. Most, I believe, are the strings from the
text stripping. Now sort by name and look at org.apache.fontbox* and
org.apache.pdfbox.*, again only at the "live" colums. At the few places where
the number was != 0, I was able to find static declarations, e.g. maps to keep
objects to speed things up. What I also found (and is related to what you
found) is that there are static maps, they are found in COSName and in PDFont.
IF MY THEORY is correct - then you might want to try to call
COSName.clearResources() and PDFont.clearResources() after strips to see if it
gets better. Of course your software will be slower.
(The sad news, for me, is that the NB profiler isn't telling the whole story -
it claims that createFont is calling PDFont<clinit> but doesn't tell the calls
between)
> PDFTextStripper.getText - hight memory usage
> --------------------------------------------
>
> Key: PDFBOX-1808
> URL: https://issues.apache.org/jira/browse/PDFBOX-1808
> Project: PDFBox
> Issue Type: Bug
> Components: Text extraction
> Affects Versions: 1.8.2, 1.8.3
> Environment: Windows 7
> Java jdk 1.7.0_45
> Reporter: Guyenot Jeremy
> Priority: Critical
> Labels: performance
> Attachments: 1808-java char copyof.jpg, 1808-java char
> copyofrange.jpg, 1808-java usage.jpg, 1808-pdfbox usage.jpg,
> 1808-snapshot.nps, DOSSIER DE CANDIDATURE_001.pdf, s5-1.png, s5-2.png,
> s50-1.png, s50-2.png
>
> Original Estimate: 72h
> Remaining Estimate: 72h
>
> Hello,
> i'm trying to extract text from pdfs but i can find that the PDFTextStripper
> use a lot of memory.
> With a pdf that have 2676 pages (for a 4.6Mo size) it use 1.5Go memory.
> I also constat that the memory is'nt free after the getText method is called.
> You can see my code bellow:
> double virgule = Math.pow(10, 2);
> System.out.println("START - Total memory (Mo): " +
> Math.round((Runtime.getRuntime().totalMemory()/1000000) * virgule) / virgule);
> PDDocument cd = PDDocument.load(file);
> System.out.println("PDDocument getNumberOfPages - Nombre de
> pages: " + cd.getNumberOfPages());
> System.out.println("PDDocument load - Total memory (Mo): " +
> Math.round((Runtime.getRuntime().totalMemory()/1000000) * virgule) / virgule);
> String pdfText = "";
> try{
> PDFTextStripper stripper = new PDFTextStripper();
> pdfText = stripper.getText(cd);
> System.out.println("PDFTextStripper getText - Total
> memory (Mo): " + Math.round((Runtime.getRuntime().totalMemory()/1000000) *
> virgule) / virgule);
> stripper.resetEngine();
> stripper = null;
> System.out.println("PDFTextStripper resetEngine - Total
> memory (Mo): " + Math.round((Runtime.getRuntime().totalMemory()/1000000) *
> virgule) / virgule);
> }
> finally{
> if( cd!=null ){
> cd.close();
> cd = null;
> System.out.println("PDDocument close - Total
> memory (Mo): " + Math.round((Runtime.getRuntime().totalMemory()/1000000) *
> virgule) / virgule);
> }
> }
> retour = new TextField(fieldName, pdfText, Field.Store.NO);
> System.out.println("TextField - Total memory (Mo): " +
> Math.round((Runtime.getRuntime().totalMemory()/1000000) * virgule) / virgule);
> And the result into my output window:
> START - Total memory (Mo): 95.0
> PDDocument getNumberOfPages - Nombre de pages: 2676
> PDDocument load - Total memory (Mo): 121.0
> PDFTextStripper getText - Total memory (Mo): 757.0
> PDFTextStripper resetEngine - Total memory (Mo): 757.0
> PDDocument close - Total memory (Mo): 757.0
> TextField - Total memory (Mo): 757.0
> pdfText - Total memory (Mo): 757.0
> I also try to call System.gc() but the memory use is the same.
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)
