Rename content fields to value

[Earlopain]

Currently there are 43 matches for name: value and only 4 for name: content in config.yml
RationalNode/ImaginaryNode also have value from node extensions.

(of course, most of the 43 are not for literals but for child nodes like in the RHS of a = 1)

The one I particularly care about is StringNode because I often handle SymbolNode and StringNode the same way but I have to write different code becaues the fields are not named the same. This allows to ducktype them and I don't have to keep in mind when to use content vs value. value for a literal also sounds better to me.

Same as in #4060, keep the previous names around.

[eregon]

Wouldn't you want unescaped instead of content for those? (depending on the situation)
I would expect unescaped if aliased to value, e.g. Integer#value parses the number, "unescaping it" for something like 1_000_000.

require 'prism'

code = <<'RUBY'
"a\nb"
RUBY

node = Prism.parse(code).value.statements.body[0]

pp node
p node.content
p node.unescaped

gives:

$ ruby test_string_node.rb
@ StringNode (location: (1,0)-(1,6))
├── flags: newline
├── opening_loc: (1,0)-(1,1) = "\""
├── content_loc: (1,1)-(1,5) = "a\\nb"
├── closing_loc: (1,5)-(1,6) = "\""
└── unescaped: "a\nb"
"a\\nb" # content
"a\nb" # unescaped

[eregon]

SymbolNode does indeed name content as value though, and has unescaped too:

code = <<'RUBY'
:"a\nb"
RUBY

@ SymbolNode (location: (1,0)-(1,7))
├── flags: newline, static_literal, forced_us_ascii_encoding
├── opening_loc: (1,0)-(1,2) = ":\""
├── value_loc: (1,2)-(1,6) = "a\\nb"
├── closing_loc: (1,6)-(1,7) = "\""
└── unescaped: "a\nb"
"a\\nb" # value
"a\nb" # unescaped

[eregon]

Or IOW, I would expect SomeLiteralNode value to return an instance of Literal, e.g. IntegerNode#value returns an Integer, FloatNode#value a Float, etc.
That's however not the case currently for SymbolNode#value which returns an escaped String, and not an unescaped Symbol.

(apologies for the many comments, just trying to explain my thoughts on this)

[Earlopain]

That did not cross my mind. Yes, I would want the unescaped one and have probably not done so many times when I should have for correctness.

Let me update this to a different approach. However I'm not so convinced anymore since the distinction between value/content is not immediatly obvious when both are present.

[kddnewton]

Yeah I'm very hesitant on this. If I were to call value on a SymbolNode, I would expect it to be the actual value of the symbol. For a string, I would expect it to be the unescaped string. Same for all other literals. Honestly I'm not too keen on this PR because it's a fair amount of churn (most clients working with Prism will touch symbols/strings) and I'm not sure it's getting us closer to a specific goal.

[Earlopain]

Not sure I'm following. What should .value be for :"a\nb" be in your mind? With the newline resolved, or just as written (so with a literal backslash + n)

[eregon]

Not sure I'm following. What should .value be for :"a\nb" be in your mind? With the newline resolved, or just as written (so with a literal backslash + n)

I believe Kevin means with the newline resolved, since it would be a Symbol object corresponding to the source.
Basically I think value for a literal node should be the same, so to speak, as eval(node.slice).
It's the case for most literals, but not SymbolNode (returns an escaped String instead), StringNode, XStringNode, RegularExpressionNode, MatchLastLineNode (no value method), and maybe a few more.

I'd imagine changing SymbolNode#value to return a Symbol instead of a String is tough for compatibility.