`flatten` separator for `keepParentNameForColumns` should not be `.`

[Jolanrensen]

Related to #666

Since we're generating accessors for columns, having a . in a column name might cause unexpected behavior. (Such as column accessors with `` around their name). _ might be a better default.

[zaleslaw]

Should we forbid a wide range of separators, or propose something from "white list" only? Let's answer ourselves

[Jolanrensen]

We should follow the kotlin spec:

• (Letter | '_') {Letter | '_' | UnicodeDigit} is allowed without `
  • Letter: any unicode character of categories Lu, Ll, Lt, Lm or Lo
  • UnicodeDigit: any unicode character of category Nd
• '`' QuotedSymbol {QuotedSymbol} '`'
  • any character excluding CR, LF and '`' (well except the last part, we cannot write ` inside a name with backticks
• ., ;, [, ], /, <, >, :, \\ are never allowed

Source: https://kotlinlang.org/spec/syntax-and-grammar.html#identifiers

That said, these never-allowed characters might appear from sources like csv/json, so I'm wondering whether we should replace them immediately or, like now, fix it only when generating accessors?

[koperagen]

I vote to keep name String as is and only replace characters in generated properties names. So, to not change this logic

[Jolanrensen]

I vote to keep name String as is and only replace characters in generated properties names. So, to not change this logic

Alright, but can the generated property then always use _ instead of ?

Seems we have some curious symbol replacements:

dataframe/core/src/main/kotlin/org/jetbrains/kotlinx/dataframe/codeGen/GeneratedField.kt

Line 117 in 0e3fdcd

public fun of(name: String): ValidFieldName {

I feel most of them could just become _ actually, that way we don't even need backticks

[koperagen]

I wouldn't want to change to _, group col looks cleaner than group_col imo. Completion even inserts backticks automatically when you type, you don't have to start with `.

[Jolanrensen]

I wouldn't want to change to _, group col looks cleaner than group_col imo. Completion even inserts backticks automatically when you type, you don't have to start with `.

It works in notebooks in the IDE yes, but I don't know about other environments. Also I've got to disagree about cleaner looks:

• df.parent_child
• df.`parent child`

I feel like backticks are a "last resort" in Kotlin. Like "okay you're not supposed to use this character, but if you really want to you can do it in backticks". If a space already exists in the name, then yes, we should just generate backticks around the name and call it a day. But I don't think we should explicitly introduce characters that need backticks to work.

[koperagen]

I wouldn't want to change to _, group col looks cleaner than group_col imo. Completion even inserts backticks automatically when you type, you don't have to start with `.

It works in notebooks in the IDE yes, but I don't know about other environments. Also I've got to disagree about cleaner looks:

• df.parent_child
• df.`parent child`

I feel like backticks are a "last resort" in Kotlin. Like "okay you're not supposed to use this character, but if you really want to you can do it in backticks". If a space already exists in the name, then yes, we should just generate backticks around the name and call it a day. But I don't think we should explicitly introduce characters that need backticks to work.

I want to make it clear that i speak about in ValidFieldName, where invalid symbol is replaced by it, and not about default flatten delimiter. group col maybe it's a confusing illustration, it's more fitting for flatten