Discussion on: Kotlin CSV to Data Class Parser

[Stephan Schroeder]

Cool, csv-parsing is a nice, self-contained but useful problem domain!
I recently wrote a csv to Kotlin data class parser for my work at 'Which?' as well 😅. I called my project KSV.

CSV contain some interesting corner cases that you might want to consider:

• column names (the comma-separated elements of the header) might contain spaces, so while the property name is a good default, you might want to add an annotation that let's you define a different name (e.g. one that contains a space)
• column names and values might be surrounded by quotes so that they can contain a comma (so your line splitting has to be a bit smarter than simply splitting by comma )
• you don't seem to support nullable types currently, it makes sense to map empty strings in the csv to null, if the type you map to is nullable
• while the separator in csv is often a comma and the quote a double quote, it's a good idea to make that configurable (e.g. because german uses a comma instead of a dot for floats, their csv often uses a semicolon instead)

advanced features:

• support conversion to LocalDateTime (you need to specify the expected format)
• support conversion to user-defined types (the user must be able to add string-to- converters)

I thing those are the main additional features I can think of right now. How did you publish to jcenter? Is it free (for open source)? I tried to google a free jar repository, but ended up suggesting Gradle's source dependency mechanism

[Victor Harlan D. Lacson]

Hi, thanks and I really appreciate the key points you shared.

• Regarding the custom delimiters, surrounded with quotes, Kotlin-Csv already handles the scenarios. That's why I mainly focused on my Kotlin-Grass as a parser itself.

• Custom Mapping is also supported on initial release, instead of relying on annotation I resorted to DSL mappings.
  

 val grass = grass<DateTimeTypes> {
        customKeyMap = mapOf("hour" to "time", "birthdate" to "date")
    }

which is I think neat and there is no need to annotate your data class

• Support on Java Date Time API conversion is also available.
• On User-defined types, I never yet encountered a common use case. I guess the user can do extension functions instead without polluting data class of unwanted logic but data itself. If it is not that complicated 😖🤣

With the release of version 0.3.0, nullable values, and white spaces were already covered.

Publishing to jcenter I never thought would be just easy, although I have done so much research since my structure is for kotlin-multiplatform. You just need to create bintray account, then publish your library in there. Once your library is uploaded you can request to link your packages to jcenter.

Apply necessary plugin in Gradle build file, like maven publish. You can check my repo's build file, it is configured for publishing in bintray.

[Stephan Schroeder]

cool, seems like you got all the features I can think of then 👍

[Ryszard Grodzicki]

I'd propose an enhancement to the customKeyMap, or implementing another parameter for stating the mapping in more kotlin way with KProperty:

val grass = grass<DateTimeTypes> {
        customKeyMap = mapOf("hour" to DateTimeTypes::time, "birthdate" to DateTimeTypes::date)
    }

Nonetheless - great job 👍

[Victor Harlan D. Lacson]

I agree with your suggestion, ill keep this in mind in the future release.

github.com/blackmo18/kotlin-grass/...