Option for null to ""

[fizerkhan]

The value 'null' seems to be irrelevant in CSV. It will be nicer if you can automatically convert to empty string "" in CSV file.

It could be even better and generic if it is customizable to any string or any number.

[zemirco]

What do you mean exactly? You have a null value in your JSON and for every null you want an empty String "" in the transformed CSV? Is that correct?

[fizerkhan]

Yes, You are correct.

[zemirco]

I don't want to implement surprising features. People, who feed in JSON with null values expect CSV with null values.

Why don't you run a find and replace on your final CSV? Look for all null values and replace them with empty Strings "".

[fizerkhan]

Yes, I can do it by either processing my input data or result(csv).
I feels, it always irrelevant to have null, undefined values in CSV file. What do you think?

[zemirco]

@fizerkhan , I won't replace null values with empty strings. I hope you understand my point of view.

[fizerkhan]

zeMirco Thanks , I understood.

[timrobertson100]

I know this is old, but rather than opening a new issue
+1 to replacing null with "".

Putting out the text "null" is both surprising and wrong in my opinion. Ideally you would be able to provide a config parameter to control what is generated for null values.

[knownasilya]

@timrobertson100 Would an option like transform, which would be a function that is passed the value, and you can transform and return the new value, be helpful for you for this scenario?

[timrobertson100]

It's cumbersome, but sure. Better would be just to be able to say "use '\N' for null" which is like MySQL, PostgreSQL etc. Sadly I can't get a CSV with nulls from this library into postgres using the native postgresql COPY loader because of the null handling.

[knownasilya]

Better would be just to be able to say "use '\N' for null" which is like MySQL, PostgreSQL etc.

That is definitely outside of the scope of this library, and much more complicated then is needed here..

[timrobertson100]

I'm sorry, but I don't understand why. It's no different to delimiter:

json2csv({ data: output.data, fields: fields, del: '|' }

Why is this so out of scope?

json2csv({ data: output.data, fields: fields, del: '|', nullVal: '\N' }

I don't mean to be awkward. It's a super nice library but doesn't produce very portable CSV files as it is.

[knownasilya]

Ah, ok I thought you were suggesting a DSL.. 😄

Yes, something like nullValue would be nice.

[timrobertson100]

Thanks @knownasilya and sorry for not just putting that example at the beginning

[knownasilya]

So that's well and good, but it won't handle undefined values, and I don't want to implement an undefinedValue as well. So I could combine them into something like unsetValue, but then what if some one wants different values for the two.

I think the best option would be a generic solution that can be used for any situation, like transform. I could add something like json2csv({ transform: json2csv.somePresetHandler }) if we can figure out some common use cases.

[timrobertson100]

Is an undefined value one where the fields references a property that is not present? If so that is by definition a null (no value for a property) and would normally be rendered the same.

Typically with JSON libraries you have the option to render nulls as propertyName: null or exclude them completely. For example in Jackson, a common Java library, you have @JsonSerialize(include = JsonSerialize.Inclusion.NON_NULL) which controls that option.

A transform() would be nice as it would allow you to tweak your own microsyntax on a cell level, but seems like it might also bring in a can of worms (e.g. the need to pass the row number, column name etc to the function).

[knownasilya]

Yes it would contain the column name and value, which isn't so much a can of worms 😄.

null and undefined are not equivalent in JS (good summary on SO), since they have different meanings for what they represent, thus some might want to render them differently.

[knownasilya]

Just remembered about the defaultValue option. Does that not work for your use case?

Described as

String, default value to use when missing data. Defaults to ('') if not specified. (Overridden by fields[].default)

In the readme

[orzarchi]

defaultValue isn't working.
I would also like to output an empty string for null cells, but the following lines (160-161) prevent it:

        if (val !== undefined) {
          var stringifiedElement = JSON.stringify(val);

when val is null, the if clause still activates and JSON.stringify converts it to the string 'null'

[knownasilya]

@orzarchi could you submit a PR with a failing test?

[knownasilya]

@orzarchi that line does not prevent it, see https://github.com/zemirco/json2csv/blob/master/lib/json2csv.js#L155 (defaultValue is used with lodashGet).

Also the tests make sure that it works here: https://github.com/zemirco/json2csv/blob/master/test/index.js#L241

Also, if val was null, it would still stringify, since that is a non-coercive equality check.

[orzarchi]

Created PR #89 with failing test. Will try to fix it as well.

[knownasilya]

Ok, I merged the test, so just work off master to get it to pass 👍 Thanks!

[knownasilya]

@orzarchi did you have a go at the fix?

[orzarchi]

yup, 20 days ago #90.

[knownasilya]

Wow, sorry I missed it. Merged. Should have a new version out tomorrow.