Tika pdf failure after upgrade to pdfBox in Camel, requires new quarkiverse-tika

[JiriOndrusek]

Bug description

Camel upgraded pdfbox to 3.x (https://issues.apache.org/jira/browse/CAMEL-19796).

Pdfbox 3.x is not back compatible with 2.x, therefore quarkiverse-tika used by tika extension fails with the new pdfbox.

Apache tika is aware of the new version of pdfbox and the upgrade ticket is already in progress - see https://issues.apache.org/jira/browse/TIKA-3347

As soon as an new version of apache tika is released it has to be adopted by quarkiverse-tika and this new version has to be adopted by camel-quarkus.

I'm disabling tika tests using pdfbox until provblem is solved (on camel-main)

[JiriOndrusek]

@ppalaga @jamesnetherton We have to wait for the new release of quarkiverse-tika supporting the new pdfbox. There is probably not much other options (the pdfbox involves i.e. fop, tika, pdf extensions, therefore keeping pdfbox on 2.x may be complicated. WDYT?

[jamesnetherton]

We have to wait for the new release quarkiverse-tika

Or we propose to revert the upgrade in Camel. I assume the same issue exists there if you try to bring the tika & pdf components together in the same app?

[JiriOndrusek]

We have to wait for the new release quarkiverse-tika

Or we propose to revert the upgrade in Camel. I assume the same issue exists there if you try to bring the tika & pdf components together in the same app?

Error starts to happen if tika parses any pdf file -> other functionality should work even with pdf extension together.

Reverting of the Camel change and postponing it until quarkiverse-tika would support 3.x would be an easy solution

[oscerd]

There is no point in postponing the change in Camel. Not for special reasons, but just because we cannot base the core development on what Quarkus/Quarkiverse does. It's dangerous and not healthy.

[oscerd]

If errors appear in Tika on plain camel, then it make sense to wait for a Tika release supporting pdfbox 3.x, but this is not evident through tests.

[JiriOndrusek]

@oscerd I'll add some tests covering pdf into tika component, because I can not see pdf file there - https://github.com/apache/camel/tree/main/components/camel-tika/src/test/resources

If some troubles emerges, we can discuss what to do next. Does it sounds ok?

[oscerd]

Yes, it is.

[oscerd]

Also if you check in the SBOM for camel, pdfbox is used explicitly only in camel-pdf and camel-fop.

Camel-tika is using only:

    {
      "ref" : "pkg:maven/org.apache.camel/[email protected]?type=jar",
      "dependsOn" : [
        "pkg:maven/org.apache.camel/[email protected]?type=jar",
        "pkg:maven/org.apache.tika/[email protected]?type=jar",
        "pkg:maven/org.apache.tika/[email protected]?type=jar",
        "pkg:maven/org.apache.tika/[email protected]?type=jar"
      ]
    }

both the tika commons and tika text modules are not using pdfbox. It's only something related to quarkiverse extension : https://github.com/quarkiverse/quarkus-tika/blob/main/pom.xml#L63

[JiriOndrusek]

I missed that fact, though having pdf part of the tika test coverage make sense nevertheless. (but with lower priority)

[JiriOndrusek]

@oscerd I tried to create a test which parses pdf (using plain Camel). For that purpose I had to add:

<dependency>
    <groupId>org.apache.tika</groupId>
    <artifactId>tika-parser-pdf-module</artifactId>
    <version>${tika-version}</version> (2.8.0)
</dependency>

to get PdfParser for Tika. This dependency brings PdfBox 2.x (see https://mvnrepository.com/artifact/org.apache.tika/tika-parser-pdf-module/2.9.0)

Therefore user which would like to use pdf parser for Tika might had a conflict in dependencies in case i.e. camel-pdf is also part of the project. The failure is caused by the no-compatibility between pdfbox 2.x and 3.x.

java.lang.NoSuchMethodError: 'org.apache.pdfbox.pdmodel.PDDocument org.apache.pdfbox.pdmodel.PDDocument.load(java.io.InputStream, java.lang.String, org.apache.pdfbox.io.MemoryUsageSetting)'
	at org.apache.tika.parser.pdf.PDFParser.getPDDocument(PDFParser.java:421)

I can imagine that this use case does not make sense - parsing pdf by camel-tika and depend on camel-pdf.

In case that useage of camel-tika (for pdf parsing) together with camel-pdf is not supported, forcing similar restriction for camel-quarkus might solve the problem. (Currently we are testing tika extension for parsing pdf files)

--- edited

tika version is 2.8. not 2.9.0 (behavior is the same)

[oscerd]

To me this is really a corner case. I understand the point, but it's a tika problem.

[JiriOndrusek]

I agree, so the right way for camel-quarkus is not testing pdf parsing with Tika (as this is the corner case in plain Camel, which might not work in some cases), Camel-pdf should be used instead.

[oscerd]

Yes, once they align we could re-use tika for parsing PDF