Spring batch ItemReader locale, import a double with comma - spring-batch

Question

I want to import the following file with Spring Batch
key;value
A;9,5
I model it with the bean
class CsvModel
{
String key
Double value
}
The shown code here is Groovy but the language is irrelevant for the problem.
#Bean
#StepScope
FlatFileItemReader<CsvModel> reader2()
{
// set the locale for the tokenizer, but this doesn't solve the problem
def locale = Locale.getDefault()
def fieldSetFactory = new DefaultFieldSetFactory()
fieldSetFactory.setNumberFormat(NumberFormat.getInstance(locale))
def tokenizer = new DelimitedLineTokenizer(';')
tokenizer.setNames([ 'key', 'value' ].toArray() as String[])
// and assign the fieldSetFactory to the tokenizer
tokenizer.setFieldSetFactory(fieldSetFactory)
def fieldMapper = new BeanWrapperFieldSetMapper<CsvModel>()
fieldMapper.setTargetType(CsvModel.class)
def lineMapper = new DefaultLineMapper<CsvModel>()
lineMapper.setLineTokenizer(tokenizer)
lineMapper.setFieldSetMapper(fieldMapper)
def reader = new FlatFileItemReader<CsvModel>()
reader.setResource(new FileSystemResource('output/export.csv'))
reader.setLinesToSkip(1)
reader.setLineMapper(lineMapper)
return reader
}
Setting up a reader is well known, what was new for me was the first code block, setting up a numberFormat / locale / fieldSetFactory and assign it to the tokenizer. However this doesn't work, I still receive the exception
Field error in object 'target' on field 'value': rejected value [5,0]; codes [typeMismatch.target.value,typeMismatch.value,typeMismatch.float,typeMismatch]; arguments [org.springframework.context.support.DefaultMessageSourceResolvable: codes [target.value,value]; arguments []; default message [value]]; default message [Failed to convert property value of type 'java.lang.String' to required type 'float' for property 'value'; nested exception is java.lang.NumberFormatException: For input string: "9,5"]
at org.springframework.batch.item.file.mapping.BeanWrapperFieldSetMapper.mapFieldSet(BeanWrapperFieldSetMapper.java:200) ~[spring-batch-infrastructure-4.1.2.RELEASE.jar:4.1.2.RELEASE]
at org.springframework.batch.item.file.mapping.DefaultLineMapper.mapLine(DefaultLineMapper.java:43) ~[spring-batch-infrastructure-4.1.2.RELEASE.jar:4.1.2.RELEASE]
at org.springframework.batch.item.file.FlatFileItemReader.doRead(FlatFileItemReader.java:180) ~[spring-batch-infrastructure-4.1.2.RELEASE.jar:4.1.2.RELEASE]
So the question is: how do I import floats in the locale de_AT (we write our decimals with a comma like this: 3,141592)? I could avoid this problem with a FieldSetMapper but I want to understand what's going on here and want to avoid the unnecessary mapper class.
And even the FieldSetMapper solution doesn't obey locales out of the box, I have to read a string and convert it myself in a double:
class PnwExportFieldSetMapper implements FieldSetMapper<CsvModel>
{
private nf = NumberFormat.getInstance(Locale.getDefault())
#Override
CsvModel mapFieldSet(FieldSet fieldSet) throws BindException
{
def model = new CsvModel()
model.key = fieldSet.readString(0)
model.value = nf.parse(fieldSet.readString(1)).doubleValue()
return model
}
}
The class DefaultFieldSet has a function setNumberFormat, but when and where do I call this function?

Answer 1

This unfortunately seems to be a bug. I have the same Problem and debugged into the code.
The BeanWrapperFieldSetMapper is not using the methods of DefaultFieldSetFactory, that would do the right conversion, but instead just uses FieldSet.getProperties and does the conversion by itself.
So, I see the following options: Provide the BeanWrapperFieldSetMapper either with PropertyEditors or a ConversionService, or use a different mapper.
Here is a sketch of a conversion Service:
private static class CS implements ConversionService {
#Override
public boolean canConvert(Class<?> sourceType, Class<?> targetType) {
return sourceType == String.class && targetType == double.class;
}
#Override
public boolean canConvert(TypeDescriptor sourceType, TypeDescriptor targetType) {
return sourceType.equals(TypeDescriptor.valueOf(String.class)) &&
targetType.equals(TypeDescriptor.valueOf(double.class)) ;
}
#Override
public <T> T convert(Object source, Class<T> targetType) {
return (T)Double.valueOf(source.toString().replace(',', '.'));
}
#Override
public Object convert(Object source, TypeDescriptor sourceType, TypeDescriptor targetType) {
return Double.valueOf(source.toString().replace(',', '.'));
}
}
and use it:
final BeanWrapperFieldSetMapper<IBISRecord> mapper = new BeanWrapperFieldSetMapper<>();
mapper.setTargetType(YourClass.class);
mapper.setConversionService(new CS());
...
new FlatFileItemReaderBuilder<IBISRecord>()
.name("YourReader")
.delimited()
.delimiter(";")
.includedFields(fields)
.names(names)
.fieldSetMapper(mapper)
.saveState(false)
.resource(resource)
.build();