Uploaded image for project: 'Tika'
  1. Tika
  2. TIKA-1606

Tika has a dependency on a very old version of Guava

    XMLWordPrintableJSON

Details

    • Improvement
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • 1.7
    • None
    • parser

    Description

      I've run into one problem while testing Tika 1.8-rc2 with Bixo

      It involves a dependency issue involving (of course) Guava, since that project loves to break their API

      The bixo-core jar has these transitive dependencies on various versions of Guava:

      Hadoop - 11.0.2
      Cascading - 14.0.1
      Tika-parsers - 10.0.1
      cdm - 17.0

      Everyone winds up using version 10.0.1 (note that Tika has a dependency on cdm, which wants to use 17.0)

      The problem is that Hadoop (for any recent version) uses an API from Guava's cache implementation that no longer exists:

      com.google.common.cache.CacheBuilder.build(Lcom/google/common/cache/CacheLoader;)Lcom/google/common/cache/LoadingCache;
      java.lang.NoSuchMethodError: com.google.common.cache.CacheBuilder.build(Lcom/google/common/cache/CacheLoader;)Lcom/google/common/cache/LoadingCache;
      at org.apache.hadoop.io.compress.CodecPool.createCache(CodecPool.java:62)
      at org.apache.hadoop.io.compress.CodecPool.<clinit>(CodecPool.java:74)
      at org.apache.hadoop.io.SequenceFile$Writer.close(SequenceFile.java:1272)
      at org.apache.hadoop.mapred.SequenceFileOutputFormat$1.close(SequenceFileOutputFormat.java:79)

      So what this means is that anyone trying to use Tika with Hadoop will need to play games with the class loader to get the older version of Guava - though that can cause other issues if Hadoop (or Cascading, etc) rely on anything that's only in the newer Guava API.

      Guava 10.0.1 was released about 3.5 years ago; 11.0.2 was from about 3 years ago. So it seems like we should upgrade to at least 11.0.2

      Attachments

        1. output.txt
          48 kB
          Lewis John McGibbney
        2. TIKA-1606.patch
          0.4 kB
          Lewis John McGibbney

        Activity

          People

            kkrugler Kenneth William Krugler
            kkrugler Kenneth William Krugler
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: