Discover the magic of the internet at Imgur, a community powered entertainment destination. Lift your spirits with funny jokes, trending memes, entertaining gifs, inspiring stories, viral videos, and so much more from users.
JSON for serialization all the way. It’s simple and to the point. It does one thing and does it well. There’s little room for annoying surprises. Any JSON can easily be minified and prettified back and forth. If you want it in binary format you can convert it to BSON.
Yaml is too much of a feature creep. It tries to do way too many things at the same time. There are so many traps to fall into if you’re not cautious enough. The same thing can be written in multitudes of ways.
There's a special place in hell for the inventor of semantically significant whitespace.
YAML itself is one of the circles of hell. You have to copy-paste YAML from web etc sources with dubious formatting for all eternity, and the editor doesn't have YAML support. Also you can only use Python
Indenting copy pasted yaml is always a pain in the butt. Any indentation you give is likely a valid yaml. Especially bad considering indentation has a significant meaning. You have to double check back and forth to ensure nothing bad has sneaked in.
With JSON there are no such discrepancies. It’s likely the editor has figured it out for you already. If it hasn’t it’s easy to prettify the JSON yourself.
Semantic whitespace problems can easily be literally impossible to solve automatically. One of the dumbest fucking ideas anybody ever came up with in computing and its inventor if anyone belongs in YAML Hell. As a fuckup it's not quite as bad as null, but that ain't exactly a high bar
I'm not sure which thought is scarier: that you don't know what you're signing up for, or that you do know and you enjoy fixing undecidable formatting fuckups manually
My problem with yaml is if you truncate it at any random spot, there's a high likelihood it's still valid yaml. I don't like the idea that things can continue without even knowing there's a problem. The single opening and closing curly braces enclosing a json object is all it takes to at least know you didn't receive the entire message. Toml has the same issue. I'll stick with json when it makes sense.
Quite like YAML, XML has too many stuff in it. While a lot of parsers are not standard compliant and safe, if there's any chance the stuff you include on your code can evolve into a fully featured parser, including it is something to avoid.
There is this language called KDL that looks interesting.
Serializing? For serializing you probably want performance above all else. I'm saying this without checking any benchmark, but I'm sure yaml is more expensive to parse than other formats where indentation don't have meaning.
For human readability: it has to be readable (and writeable) by all humans. I know (a lot of people) that dislike yaml, toml and XML. I don't know of a single person that struggles to read/write json, there is a clear winner.
JSON would be perfect if it allowed for comments. But it doesn't and that alone is enough for me to prefer YAML over JSON. Yes, JSON is understandable without any learning curve, but having a learning curve is not always bad. YAML provides a major benefit that is worth the learning curve and doesn't have the issues that XML has (which is that there is no way to understand an XML without also having the XSD for it)
If a comment isn't part of the semantic content of a JSON object it has no business being there. JSON models data, it's not markup language for writing config files.
For the data interchange format, comments aren't part of the JSON grammar but the option to parse non-JSON values is left open to the implementation. Many implementations do detect (and ignore) comments indicated by e.g. # or //.
I don't know why we're fucking about trying to use text editors to manipulate structured data.
Yeah, it's convenient to just be able to use a basic text editor, but we're not trying to cram it all on a floppy disk here. I'm sure we could have a nice structured data editor somewhere for all those XML, JSON and YAML files we're supposed to maintain every day.
Serializing isn't necessarily about performance, or we'd just use protobuf or similar. I agree Json is a great all rounder. Combine with JSON object schema to define sophisticated DSLs that are still readable, plain JSON. TOML is nice as a configuration language, but its main appeal (readability) suffers when applied to complex modeling tasks. XML is quite verbose and maybe takes the "custom DSL" idea a little too far. YAML is a mistake.
Ever tried NestedText? It's like basic YAML but everything is a string (types are up to the code that ingests it), and you never ever need to escape a character.
I've got too many consumers that I don't control which dictate their input formats. And to be quite honest, "types are up to the code that ingests it" sounds like a huge negative to me.