The complaints about yaml’s quirks (no evaluating to false, implicit strings, weird number formats, etc.) are valid in theory but I’ve never encountered them causing any real-life issues.
no doesn’t become false, it becomes Norway, and when converted to a boolean, Norway is true. The reason’s because one on YAML’s native types is an ISO country code enum, and if you tell a compliant YAML implementation to load a file without giving it a schema, that type has higher priority than string. If you then call a function that converts from native type to string, it expands the country code to the country name, and a function that coerces to boolean makes country codes true.
The problem’s easy to avoid, though. You can just specify a schema, or use a function that grabs a string/bool directly instead of going via the assumed type first.
The real problem with YAML is how many implementations are a long way from being conformant, and load things differently to each other, but that situation’s been improving.
Are you sure? I’ve always heard it the other way around and a quick search for "YAML norway’ gives this
The reason to why this is problematic in some cases, is “The Norway Problem” YAML has: when you abbreviate Norway to its ISO 3166-1 ALPHA-2 form NO, YAML will return false when parsing it
Also, YAML 1.2 (2009) changed the format of booleans to only be case insensitive true and false. “No” no longer is false if you’re parsing as a version 1.2 document.
Yeah, looks like I’d remembered it backwards. It’s still an easily solvable problem by not using a load everything as whatever type you feel like function.
The complaints about yaml’s quirks (
noevaluating tofalse, implicit strings, weird number formats, etc.) are valid in theory but I’ve never encountered them causing any real-life issues.nodoesn’t becomefalse, it becomesNorway, and when converted to a boolean, Norway is true. The reason’s because one on YAML’s native types is an ISO country code enum, and if you tell a compliant YAML implementation to load a file without giving it a schema, that type has higher priority than string. If you then call a function that converts from native type to string, it expands the country code to the country name, and a function that coerces to boolean makes country codes true.The problem’s easy to avoid, though. You can just specify a schema, or use a function that grabs a string/bool directly instead of going via the assumed type first.
The real problem with YAML is how many implementations are a long way from being conformant, and load things differently to each other, but that situation’s been improving.
That is somehow so much worse
I believe they’re getting themselves confused.
nowas false prior to YAML 1.2. This is known as the “Norway problem.”Are you sure? I’ve always heard it the other way around and a quick search for "YAML norway’ gives this
Also, YAML 1.2 (2009) changed the format of booleans to only be case insensitive true and false. “No” no longer is false if you’re parsing as a version 1.2 document.
Yeah, looks like I’d remembered it backwards. It’s still an easily solvable problem by not using a load everything as whatever type you feel like function.