blog: Add 2025-07-26.md

Author: ingydotnet

blog/2025-07-26.md

@@ -0,0 +1,107 @@
+---
+title: The YS Core YAML-Schema
+date: 2025-07-26
+draft: false
+authors: [ingydotnet]
+categories: [Summer-of-YS]
+edit: blog/2025-07-26.md
+comments: true
+---
+
+The term "YAML Schema" is a bit of unfortunate history.
+
+It's not the YAML version of a JSON Schema.
+
+In my opinion it needs to be renamed because it's not a schema in the typical
+sense of the word.
+The term has been part of the YAML spec since the beginning.
+
+Let's use the term "YAML-Schema" instead of "YAML Schema" to be clear that this
+is not a typical schema applied to a YAML file.
+
+Typically a schema is a set of rules that are used to validate data.
+
+A YAML-Schema is a set of rules about how untagged nodes are implicitly tagged
+during the loading process.
+
+Allow me to explain.
+
+<!-- more -->
+
+
+## YAML Implicit Tagging
+
+We've covered this before, but when YAML is loaded, every node without a tag is
+assigned one based on "how it looks".
+
+By default, mappings, sequences and not-unquoted-scalars are assigned the tags
+`!!map`, `!!seq`, and `!!str` respectively.
+
+In a typical loader, every tag must have been configured to identify a
+constructor function.
+The constructor phase is the final phase of the loading process.
+It creates the value that the program doing the loading will see.
+
+So what's left?
+
+It's the plain (unquoted) scalar nodes.
+
+In JSON, the allowed unquoted scalars are: `true`, `false`, `null` and scalars
+that look like numbers.
+
+A YAML-Schema defines the patterns that unquoted scalars must match, and the
+associated tag they will be assigned.
+
+
+## The YS Core YAML-Schema
+
+YS tries to stick close to the YAML 1.2 spec's [Core Schema](
+https://yaml.org/spec/1.2.2/#103-core-schema).
+This is supposed to be what most YAML Loaders should use by default.
+Therefore it should be the most compatible with other YAML Loaders.
+
+I'll show the different schema tags as actual `ys` commands:
+
+```bash
+# !!null
+$ ys -p - <<<'"!!null": [null, Null, NULL, ~]'
+{"!!null" [nil nil nil nil]}
+
+# !!bool
+$ ys -p - <<<'"!!bool": [true, True, TRUE, false, False, FALSE]'
+{"!!bool" [true true true false false false]}
+
+# !!int
+$ ys -p - <<<'"!!int": [-42, -0, 0, 42, 0x123, 0o123]'
+{"!!int" [-42 0 0 42 291 83]}
+
+# !!float
+$ ys -p - <<<'"!!float": [.123, 1.23, 1.2e3, -1.2e-3, 0.0, -0.0]'
+{"!!float" [0.123 1.23 1200.0 -0.0012 0.0 -0.0]}
+
+# Unsupported from YAML 1.2 Core Schema spec
+$ ys -p - <<<'"!!float": [.inf, .Inf, .INF, -.inf, -.Inf, -.INF, .nan, .Nan, .NAN]'
+Compile error: Inf and NaN not supported in YS
+
+# !!str for everything else
+$ ys -p - <<<'"!!str": [NulL, FaLsE, 42 + 1, 1.2.3]'
+{"!!str" ["NulL" "FaLsE" "42 + 1" "1.2.3"]}
+```
+
+
+## Using Other YAML-Schemas
+
+The above is the default YAML-Schema for YS.
+
+YAML was always meant to allow you to define your own YAML-Schema.
+
+Some implementations of YAML let you choose from a few different YAML-Schemas.
+
+But really you should be able to trivially define your own YAML-Schema for any
+YAML file you want to load.
+
+YAML never really lived up to this promise.
+
+YS does not yet support this, but it's on the roadmap.
+
+Stay tuned. Stay YS!