Xbase++ Requires Chilkat v11.7.0+

Assess PDF Text Quality Before Extracting

See more PDF Text Extraction Examples

Demonstrates the "assess" operation, which returns only the quality report: whether the text can be trusted, which pages are scanned images that need OCR, and how each page's text was found. Use it to triage documents before extracting. (Chilkat does not perform OCR.) Xbase++

LOCAL nSuccess
LOCAL oPdf
LOCAL oJson
LOCAL oJsonOut
LOCAL j
LOCAL nCount_j
LOCAL i
LOCAL nCount_i

nSuccess := 0

//  This example requires the Chilkat API to have been previously unlocked.
//  See Global Unlock Sample for sample code.

//  The PDF used in this example (page 2 is a scanned image with no text) is available at mixed_text_image.pdf

oPdf := CreateObject("Chilkat.Pdf")
nSuccess := oPdf:LoadFile("qa_data/pdf/mixed_text_image.pdf")
IF (nSuccess == 0)
    ? oPdf:LastErrorText
    oPdf:destroy()
    RETURN
ENDIF

oJson := CreateObject("Chilkat.JsonObject")
oJsonOut := CreateObject("Chilkat.JsonObject")

//  "assess" computes the quality report without producing any text.
//  It is fast, and it is always allowed, even for an encrypted PDF that forbids copying.
nSuccess := oPdf:ExtractText("assess", oJson, oJsonOut)
IF (nSuccess == 0)
    ? oPdf:LastErrorText
    oPdf:destroy()
    oJson:destroy()
    oJsonOut:destroy()
    RETURN
ENDIF

//  The overall rating is the worst rating among the pages that have text:
//  "good", "fair", "poor", or "none" when there is no text at all.
? "Rating: " + oJsonOut:StringOf("quality.rating")

//  Is the PDF tagged (does it have a structure tree)?
? "Tagged: " + Str(oJsonOut:BoolOf("quality.tagged"))

//  The pages that are images only and would need OCR, such as "2" or "4,9".
? "Pages needing OCR: " + oJsonOut:StringOf("quality.pagesNeedingOcr")

//  The per-page report.

i := 0
nCount_i := oJsonOut:SizeOfArray("quality.pages")
DO WHILE i < nCount_i
    oJsonOut:I := i
    //  textSource is "text" (real text), "ocrLayer" (invisible text over an image),
    //  "imageOnly" (needs OCR), "mixed", or "empty".
    ? "Page " + Str(oJsonOut:IntOf("quality.pages[i].page")) + ": rating " + oJsonOut:StringOf("quality.pages[i].rating") + ", textSource " + oJsonOut:StringOf("quality.pages[i].textSource")
    //  Each issue has a machine-readable code and a sentence.
    j := 0
    nCount_j := oJsonOut:SizeOfArray("quality.pages[i].issues")
    DO WHILE j < nCount_j
        oJsonOut:J := j
        ? "    " + oJsonOut:StringOf("quality.pages[i].issues[j].code") + ": " + oJsonOut:StringOf("quality.pages[i].issues[j].message")
        j := j + 1
    ENDDO
    i := i + 1
ENDDO

//  Decide what to do with the document:
IF (oJsonOut:StringOf("quality.rating") == "none")
    ? "No extractable text; the whole document needs OCR."
ELSE
    IF (oJsonOut:StringOf("quality.pagesNeedingOcr") != "")
        ? "Some pages need OCR; the remaining pages can be extracted with the text or markdown operation."
    ELSE
        ? "All pages can be extracted."
    ENDIF

ENDIF

//  Sample output:

//  Rating: good
//  Tagged: false
//  Pages needing OCR: 2
//  Page 1: rating good, textSource text
//  Page 2: rating none, textSource imageOnly
//      imageOnly: No text; page is a scanned image
//  Page 3: rating good, textSource text
//  Some pages need OCR; the remaining pages can be extracted with the text or markdown operation.

oPdf:destroy()
oJson:destroy()
oJsonOut:destroy()