feat: decode v4.* notebooks

Decoder for v4.4 is reused for all notebooks with major version 4, because they
their differences do not affect how the notebooks are rendered:
- v4.5 requires that each cell has a unique ID
- versions < v4.3 do not have the 'code_cell.metadata.execution' field, which holds the code's execution time.
  They also do not have 'raw_cell.metadata.jupyterf.source_hidden' which controls if the source is hidden.
  This has a default behaviour in 'nb' and is probably not that important anyways.
  Finally, they miss metadata.title field, which is currently not used also in v4.4 notebooks.

decode/decode_test.go

@@ -37,6 +37,16 @@ func TestDecodeBytes(t *testing.T) {
 			json   string
 			nCells int
 		}{
+			{
+				name: "v4.5",
+				json: `{
+					"nbformat": 4, "nbformat_minor": 5, "metadata": {}, "cells": [
+						{"id": "a", "cell_type": "markdown", "metadata": {}, "source": []},
+						{"id": "b", "cell_type": "markdown", "metadata": {}, "source": []}
+					]
+				}`,
+				nCells: 2,
+			},
 			{
 				name: "v4.4",
 				json: `{
@@ -47,6 +57,46 @@ func TestDecodeBytes(t *testing.T) {
 				}`,
 				nCells: 2,
 			},
+			{
+				name: "v4.3",
+				json: `{
+					"nbformat": 4, "nbformat_minor": 3, "metadata": {}, "cells": [
+						{"cell_type": "markdown", "metadata": {}, "source": []},
+						{"cell_type": "markdown", "metadata": {}, "source": []}
+					]
+				}`,
+				nCells: 2,
+			},
+			{
+				name: "v4.2",
+				json: `{
+					"nbformat": 4, "nbformat_minor": 2, "metadata": {}, "cells": [
+						{"cell_type": "markdown", "metadata": {}, "source": []},
+						{"cell_type": "markdown", "metadata": {}, "source": []}
+					]
+				}`,
+				nCells: 2,
+			},
+			{
+				name: "v4.1",
+				json: `{
+					"nbformat": 4, "nbformat_minor": 1, "metadata": {}, "cells": [
+						{"cell_type": "markdown", "metadata": {}, "source": []},
+						{"cell_type": "markdown", "metadata": {}, "source": []}
+					]
+				}`,
+				nCells: 2,
+			},
+			{
+				name: "v4.0",
+				json: `{
+					"nbformat": 4, "nbformat_minor": 0, "metadata": {}, "cells": [
+						{"cell_type": "markdown", "metadata": {}, "source": []},
+						{"cell_type": "markdown", "metadata": {}, "source": []}
+					]
+				}`,
+				nCells: 2,
+			},
 		} {
 			t.Run(tt.name, func(t *testing.T) {
 				nb, err := decode.Bytes([]byte(tt.json))

schema/schema.go

@@ -1,3 +1,11 @@
+// Package schema defines the common data format for elements of a Jupyter notebook.
+//
+// It is based on the [v4.4] definition, as it is stable and encompasses all the data
+// necessary for accurate rendering. Note, that schema validation is not a goal of this
+// package, and so, interfaces defined here will often omit the non-essential data,
+// e.g. metadata or fields specific to JupyterLab environment. 
+//
+// [v4.4]: https://github.com/jupyter/nbformat/blob/main/nbformat/v4/nbformat.v4.4.schema.json
 package schema
 
 import (
@@ -34,6 +42,9 @@ type Cell interface {
 	Text() []byte
 }
 
+// HasAttachments is implemented by cells which include [cell attachments].
+//
+// [cell attachments]: https://nbformat.readthedocs.io/en/latest/format_description.html#cell-attachments
 type HasAttachments interface {
 	// Attachments are only defined for v4.0 and above for markdown and raw cells
 	// and may be omitted in the JSON. Cells without attachments should return nil.

schema/v4/schema.go

@@ -11,13 +11,21 @@ import (
 )
 
 func init() {
-	decode.RegisterDecoder(version, new(decoder))
+	decode.RegisterDecoder(schema.Version{Major: 4, Minor: 5}, new(decoder))
+	decode.RegisterDecoder(schema.Version{Major: 4, Minor: 4}, new(decoder))
+	decode.RegisterDecoder(schema.Version{Major: 4, Minor: 3}, new(decoder))
+	decode.RegisterDecoder(schema.Version{Major: 4, Minor: 2}, new(decoder))
+	decode.RegisterDecoder(schema.Version{Major: 4, Minor: 1}, new(decoder))
+	decode.RegisterDecoder(schema.Version{Major: 4, Minor: 0}, new(decoder))
 }
 
-var version = schema.Version{Major: 4, Minor: 4}
-
+// decoder decodes cell contents and metadata for nbformat v4.4.
+// Other versions can be decoded using the same, as their schema
+// differs in ways that does not affect how the notebook is rendered.
 type decoder struct{}
 
+var _ decode.Decoder = (*decoder)(nil)
+
 func (d *decoder) DecodeMeta(data []byte) (schema.NotebookMetadata, error) {
 	var nm NotebookMetadata
 	if err := json.Unmarshal(data, &nm); err != nil {