Hello! Thanks for all your work on gephi.

We recently upgraded our project from gephi-toolkit 0.8 to gephi-toolkit 0.9.6. In doing so one of our tests that previously passed started failing. The test imported a simple VNA file, and nothing changed with that: same test file, same test code. I've pared down the import code to the below and confirmed the issue happens without any of our project's code involved.

This seems like a bug to me, but if I'm missing something that changed between 0.8 and 0.9 that would require a change to our code to keep these imports working, please let me know.

Here is the full.vna file contents:

*Node data
ID tvar breed spots fur missing
1 true MICE "" white foo
2 false FROGS true "" foo
3 false FROGS true "" foo
*Node properties
ID x y size color shortlabel
1 -355.93912 -114.81057 10.0 153 1
2 -170.28754 214.69585 10.0 153 2
3 526.2267 -99.88528 10.0 153 3
*Tie data
from to strength breed lvar
1 2 1.0 directed-edges 5.0
3 1 10.0 undirected-edges ""
1 3 10.0 undirected-edges ""

Here is the Scala code for the import.

import java.io.{ File, FileReader }

import scala.collection.JavaConverters._

import org.gephi.io.importer.api.ImportController
import org.openide.util.Lookup

object GephiVnaBugPoc {
  def main(args: Array[String]) = {
    val importController = Lookup.getDefault.lookup(classOf[ImportController])
    val file      = new File("full.vna")
    val reader    = new FileReader(file)
    val importer  = importController.getFileImporter(file)
    val container = importController.importFile(reader, importer)
    val unloader  = container.getUnloader
    val report    = container.getReport
    val nodes     = unloader.getNodes
    val nodeInfo  = nodes.asScala.map( (node) => {
      val attributes = node.getColumns.asScala.map( (column) => {
        val value = node.getValue(column.getId)
        s"(${column.getId} -> $value)"
      })
      s"${node.getId}: ${attributes.mkString(" ")}"
    }).mkString("\n")

    println("* Import report:\n")
    println(report.getText(true))
    println("* Nodes:\n")
    println(nodeInfo)
  }
}

And here is the output when that is run:

Jul 28, 2022 10:33:45 AM org.netbeans.modules.masterfs.watcher.Watcher getNotifierForPlatform
INFO: Native file watcher is disabled
* Import report:

Node column 'tvar' (String)
Node column 'breed' (String)
Node column 'spots' (String)
Node column 'fur' (String)
Node column 'missing' (String)
Value can't be null

* Nodes:

1: (tvar -> true) (breed -> MICE) (spots -> null) (fur -> null) (missing -> null)

So it appears when the first node line is read 1 true MICE "" white foo that "" value is causing a null value to be generated, which is then rejected. This also stops the processing of the nodes, as only that one is imported. I didn't add it to the output, but no edges are imported either.

Possible Solution

I didn't fully explore the gephi code, but it seemed like something might be off with how the regex is used in the ImporterVNA.split() method.

Your Environment

• Version used: Gephi 0.9.6
• Operating System: I'm running on Ubuntu 20.04 through WSL using Java 17.