CSVDecoder: No Long and Int out of range exceptions

[Pavel38l]

I have a problem when I try to process a CSV file entered by the user.
If the numeric values ​​are outside the limits of Int or Long, then there is no way I can notify the user about incorrect values ​​that are too large.

Values ​​are corrupted by _getBigInteger().intValue() and _getBigInteger().longValue() in CsvDecoder
And no error occurs
Is it possible to use the methods _getBigInteger().intValueExact() and _getBigInteger().longValueExact() which check boundaries and cause an error?

Example (Kotlin):
CSV file:

testInt,testLong
111111111111111111111111111111111111111111,2222222222222222222222222222222222222222

Entity:

import com.fasterxml.jackson.annotation.JsonProperty

open class TestEntity(
        @JsonProperty("testInt")  val testInt: Int,
        @JsonProperty("testLong") val testLong: Long,
)

import com.fasterxml.jackson.dataformat.csv.CsvMapper
import com.fasterxml.jackson.dataformat.csv.CsvParser
import java.io.File
import java.io.InputStream

class TestClass {
    fun readConvertFile() {
        val file = File("example.csv")
        val entities = readCsv(TestEntity::class.java, file.inputStream())
        println(entities[0].testInt)
        println(entities[0].testLong)
    }

    companion object {
        private val csvMapper = CsvMapper()
            .enable(CsvParser.Feature.SKIP_EMPTY_LINES)
            .enable(CsvParser.Feature.TRIM_SPACES)
            .enable(CsvParser.Feature.WRAP_AS_ARRAY)
            .enable(CsvParser.Feature.INSERT_NULLS_FOR_MISSING_COLUMNS)

        fun <T> readCsv(classArg: Class<T>, stream: InputStream): List<T> {
            val schema = csvMapper.schemaFor(classArg).withHeader().withColumnReordering(true).withNullValue("")
            val reader = csvMapper.readerFor(classArg).with(schema)
            return reader.readValues<T>(stream).readAll()
        }
    }
}

ER: some error message like jackson ObjectMapper Numeric value (111111111111111111111111111111111111111111) out of range of int
AR: Values corrupted

-954437177
-1121560256615750770

I use

<dependency>
    <groupId>com.fasterxml.jackson.dataformat</groupId>
    <artifactId>jackson-dataformat-csv</artifactId>
    <version>2.17.2</version>
</dependency>

[cowtowncoder]

Quick question: which Jackson version is this with? If not 2.17(.2), then would be good to verify issue with 2.17.2.
If 2.17.2, there is 2.18.0-SNAPSHOT -- there have been changes to number conversions.

Aside from that: just to make sure -- you would like to get an appropriate exception for value overflow wrt int/Integer and long/Long? If so, makes sense.

[Pavel38l]

@cowtowncoder hi
I used

<dependency>
    <groupId>com.fasterxml.jackson.core</groupId>
    <artifactId>jackson-databind</artifactId>
    <version>2.17.2</version>
</dependency>

and jackson-dataformat-csv: 2.17.2
Version 2.18.0 is not yet available in the central Maven repository, I don't know how to check it
https://mvnrepository.com/artifact/com.fasterxml.jackson.core/jackson-databind
In addition, the CSVDecoder class branch 2.18 contains the same methods that I listed above

Yes, that's right, I would like to receive an error about the overflow of Int and Long values

[cowtowncoder]

2.18.0 would be 2.18.0-SNAPSHOT (having Sonatype OSS Repo configured). But as you say, it is unlikely to differ in behavior.

[cowtowncoder]

Ok, was able to reproduce the issue; adding failing test via #487.

[cowtowncoder]

Fixed, to be include in 2.18.0.

Oddly enough, coercion limits were enforced if converting from floating-point values.
And the reason this was missing was because it's not CsvParser that tracks number values, conversions, but CsvDecoder -- which cannot rely on ParserBase that has necessary checks. So code is copy-pasted.