9

I have a text file like this:

ids.txt

1000
999
745
123
...

I want to read this file and load it in a two dimensional array. I expect to have an array similar to the one below:

Object[][] data = new Object[][] { //
     { new Integer(1000) }, //
     { new Integer(999) }, //
     { new Integer(745) }, //
     { new Integer(123) }, //
     ...
};

Here is the code I wrote:

File idsFile = ... ;
try (Stream<String> idsStream = Files.lines(idsFile.toPath(), StandardCharsets.US_ASCII)) {
    Object[][] ids = idsStream
       .filter(s -> s.trim().length() > 0)
       .toArray(size -> new Object[size][]);

    // Process ids array here...
}

When running this code, an exception is raised:

java.lang.ArrayStoreException: null
at java.lang.System.arraycopy(Native Method) ~[na:1.8.0_45]
at java.util.stream.SpinedBuffer.copyInto(Unknown Source) ~[na:1.8.0_45]
at java.util.stream.Nodes$SpinedNodeBuilder.copyInto(Unknown Source) ~[na:1.8.0_45]
at java.util.stream.SpinedBuffer.asArray(Unknown Source) ~[na:1.8.0_45]
at java.util.stream.Nodes$SpinedNodeBuilder.asArray(Unknown Source) ~[na:1.8.0_45]
at java.util.stream.ReferencePipeline.toArray(Unknown Source) ~[na:1.8.0_45]
...

How can resolve this exception?

Tagir Valeev
  • 97,161
  • 19
  • 222
  • 334
Stephan
  • 41,764
  • 65
  • 238
  • 329

2 Answers2

16

Given a Stream<String> you can parse each item to an int and wrap it into an Object[] using:

 strings
        .filter(s -> s.trim().length() > 0)
        .map(Integer::parseInt)
        .map(i -> new Object[]{i})

Now to turn that result into a Object[][] you can simply do:

Object[][] result = strings
        .filter(s -> s.trim().length() > 0)
        .map(Integer::parseInt)
        .map(i -> new Object[]{i})
        .toArray(Object[][]::new);

For the input:

final Stream<String> strings = Stream.of("1000", "999", "745", "123");

Output:

[[1000], [999], [745], [123]]
Boris the Spider
  • 59,842
  • 6
  • 106
  • 166
  • 5
    If you suspect that strings may sporadically have spaces before and after, then it's better to add the separate `trim` step: `strings.map(String::trim).filter(s -> !s.isEmpty())`. Also I would use `Integer::valueOf` instead of `Integer::parseInt` as it matches better the wanted type (`Integer`, not `int`). Though probably that's a matter of taste. – Tagir Valeev Jun 13 '15 at 10:44
  • @TagirValeev Why is it "better to add the separate trim step" ? – Stephan Jun 13 '15 at 10:56
  • 2
    @Stephan: this way you can parse files which have spaces after numbers (like `"1 \n2\n3 \n"`). Currently you will have `NumberFormatException` on such files. – Tagir Valeev Jun 13 '15 at 10:57
  • Do you think there is a benefit in these two `map` operations instead of simply saying `.map(s -> new Object[]{Integer.parseInt(s)})`? – Holger Jun 15 '15 at 09:23
  • @Holger don't know really. I like to split each logical transformation into a separate `map` step - I think it makes to code more readable. I suppose it's just a matter of personal preference - I doubt there's much of a performance impact. – Boris the Spider Jun 15 '15 at 09:24
5

Your last line should probably be size -> new Object[size], but you would need to provide arrays of Integers of size one and you would also need to parse the strings into Integers.

I suggest the following:

try (Stream<String> idsStream = Files.lines(idsFile.toPath(), StandardCharsets.US_ASCII)) {
    Object[][] ids = idsStream
       .map(String::trim)
       .filter(s -> !s.isEmpty())
       .map(Integer::valueOf)
       .map(i -> new Integer[] { i })
       .toArray(Object[][]::new);

    // Process ids array here...
}
assylias
  • 321,522
  • 82
  • 660
  • 783