Data

Every visualization block needs data. You have three options, in order of simplicity.

1. Inline CSV (default)

The fence body is a CSV with a header row.

```chart type=bar x=region y=sales
region, sales
North, 120
South, 95
```

Parsing:

• Comma-separated. Whitespace around values is trimmed.
• Quoted fields handle commas and embedded quotes: "Smith, J.", "He said ""hi""".
• Types auto-detected per cell: numbers → number, true/false/null → their literals, everything else → string.
• No type declarations are needed or supported.

2. Inline JSON

Opt in with format=json. The body must be a JSON array of objects.

```chart type=pie label=name value=share format=json
[
  {"name": "Alpha", "share": 45},
  {"name": "Beta",  "share": 30}
]
```

Use JSON when your data is already structured that way (e.g. pasted from an API response you've pre-aggregated). CSV is more convenient for most typing-by-hand cases.

3. Named datasets (file references)

Declare once in front-matter, reuse everywhere.

---
data:
  sales: ./data/sales.csv
  regions: ./data/regions.json
  q1: ./data/q1.tsv
---

Then reference by name:

```chart type=line data=sales x=month y=revenue series=region
```

```table data=sales
```

File resolution rules:

• Paths are always relative to the .mdv file.
• Absolute paths and URLs are not supported in v1 — a hard line to keep docs portable.
• Supported extensions: .csv, .tsv (parsed like CSV), .json.
• Each file is read once per render and cached by absolute path.
• Missing or unreadable files render an inline error banner but the rest of the document continues.

Error cases

Situation	What happens
Unknown data= name	Block renders as a red error banner; other blocks still render.
Missing file	Block renders as a red error banner naming the path.
Malformed CSV	Block renders as a red error banner.
Empty dataset	Block renders as a red "no data rows" banner.
Column referenced by x=/y= that doesn't exist	Treated as undefined → numerical cells become NaN; usually visible as blank bars/flat lines.

What's not supported

MDV v1 is deliberately small. These are not data features:

• No transformations. No filter, group-by, aggregate, compute, or sort. Pre-aggregate your data.
• No URLs or APIs. Only local files.
• No queries. You can't say data=sales where region=North. That's code, which v1 explicitly avoids.
• No parametrization. No {{ }} templating of info strings with data values.

If you find yourself wanting these, pre-process the data in your tool of choice and save the result as a new CSV/JSON.