# tiptap-rusty-parser
Fast, schema-agnostic parser & manipulator for [Tiptap](https://tiptap.dev) /
ProseMirror `JSONContent` documents, in Rust.
- **Schema-agnostic** — any node/mark `type` is accepted; unknown JSON fields are
preserved for lossless roundtrip.
- **Query** with predicate closures — `find`, `find_all`, `walk`, `descendants`.
- **Mutate** in place — marks, attrs, children, text, and bulk `replace_all`.
- **Build** ergonomically — `Node::element`, `Node::text`, `doc(..)`, `with_*` chaining.
- **Fast** — borrow over copy, stack-based traversal (no recursion blowup), `lto`
release profile, criterion benches.
---
## Table of contents
- [Install](#install)
- [Quick start](#quick-start)
- [Data model](#data-model)
- [Parsing & serializing](#parsing--serializing)
- [Querying](#querying)
- [Mutating](#mutating)
- [Marks](#marks)
- [Attributes](#attributes)
- [Children](#children)
- [Text](#text)
- [Bulk transforms](#bulk-transforms)
- [Building nodes](#building-nodes)
- [Error handling](#error-handling)
- [Performance](#performance)
- [Development](#development)
- [License](#license)
---
## Install
Add to `Cargo.toml`:
```toml
[dependencies]
tiptap-rusty-parser = "0.1"
```
Requires a recent stable Rust (edition 2021).
---
## Quick start
```rust
use tiptap_rusty_parser::{Document, Mark, Node};
fn main() -> Result<(), tiptap_rusty_parser::ParseError> {
let mut doc = Document::from_json_str(
r#"{"type":"doc","content":[
{"type":"paragraph","content":[{"type":"text","text":"hi"}]}
]}"#,
)?;
// Bold every text node.
doc.replace_all(
|n| n.node_type.as_deref() == Some("text"),
|n| { n.add_mark(Mark::new("bold")); },
);
// Append a new paragraph.
doc.push_child(Node::element("paragraph").with_text("bye"));
let json = doc.to_json_str()?;
println!("{json}");
Ok(())
}
```
---
## Data model
A Tiptap document is a tree of nodes. The crate mirrors Tiptap's `JSONContent`
shape directly.
```rust
pub struct Node {
pub node_type: Option<String>, // JSON "type", e.g. "doc", "paragraph", "text"
pub attrs: Option<Map<String, Value>>,// node attributes
pub content: Option<Vec<Node>>, // child nodes
pub marks: Option<Vec<Mark>>, // marks (bold, italic, link, …)
pub text: Option<String>, // text payload (text nodes)
pub extra: Map<String, Value>, // any unknown top-level fields, preserved
}
pub struct Mark {
pub mark_type: String, // JSON "type", e.g. "bold"
pub attrs: Option<Map<String, Value>>,// mark attributes (e.g. link href)
pub extra: Map<String, Value>, // unknown fields, preserved
}
```
`Map`/`Value` are re-exported from `serde_json`. The crate is built with the
`preserve_order` feature so attribute key order survives a roundtrip.
**Why everything is `Option`** — to faithfully distinguish *missing* from
*empty*. `content: None` serializes to no `content` key; `content: Some(vec![])`
serializes to `"content": []`. Unknown node types (custom Tiptap extensions) and
unknown fields land in `extra` and roundtrip untouched.
`Document` is a thin owning wrapper around the root `Node` and **derefs to it**,
so every `Node` method below is also callable directly on a `Document`.
---
## Parsing & serializing
```rust
use tiptap_rusty_parser::Document;
use serde_json::json;
// From a JSON string
let doc = Document::from_json_str(r#"{"type":"doc","content":[]}"#)?;
// From a serde_json::Value
let doc = Document::from_value(json!({ "type": "doc", "content": [] }))?;
// From any reader (file, socket, …)
let file = std::fs::File::open("doc.json")?;
let doc = Document::from_reader(file)?;
// Serialize
let compact = doc.to_json_str()?; // String, compact
let pretty = doc.to_string_pretty()?; // String, indented
let value = doc.to_value()?; // serde_json::Value
# Ok::<(), tiptap_rusty_parser::ParseError>(())
```
Roundtrip is lossless — unknown node types, extra fields, and key order are all
preserved.
Access the root node explicitly when needed: `doc.root()`, `doc.root_mut()`,
`doc.into_root()`. Wrap an existing node with `Document::new(node)` or
`node.into()`.
---
## Querying
All traversal is **depth-first pre-order** (a node is visited before its
children). Selection is done with predicate closures — no selector DSL to learn.
```rust
use tiptap_rusty_parser::{Document, Node};
let doc = Document::from_json_str(r#"{
"type":"doc","content":[
{"type":"paragraph","content":[
{"type":"text","text":"a","marks":[{"type":"bold"}]},
{"type":"text","text":"b"}
]}
]}"#)?;
// First match (incl. the node itself)
let first_para: Option<&Node> = doc.find(|n| n.node_type.as_deref() == Some("paragraph"));
// All matches
let texts: Vec<&Node> = doc.find_all(|n| n.node_type.as_deref() == Some("text"));
assert_eq!(texts.len(), 2);
// Predicate can inspect anything: marks, attrs, text…
// Lazy iterator over self + all descendants
let count = doc.descendants().count();
// Visit every node
let mut n = 0;
|--------|-----------|---------|
| `find` | `find(\|&Node\| -> bool)` | `Option<&Node>` |
| `find_mut` | `find_mut(\|&Node\| -> bool)` | `Option<&mut Node>` |
| `find_all` | `find_all(\|&Node\| -> bool)` | `Vec<&Node>` |
| `find_all_mut` | `find_all_mut(&mut \|&Node\| -> bool)` | `Vec<&mut Node>` |
| `walk` | `walk(&mut \|&Node\|)` | `()` |
| `walk_mut` | `walk_mut(&mut \|&mut Node\|)` | `()` |
| `descendants` | `descendants()` | `impl Iterator<Item = &Node>` |
---
## Mutating
Mutation is **in place** on a `&mut Node` / `&mut Document` — no copies, no
rebuild. Container fields auto-collapse to `None` when they become empty (e.g.
removing the last mark sets `marks` back to `None`), keeping output clean.
### Marks
```rust
use tiptap_rusty_parser::{Mark, Node};
let mut t = Node::text("hello");
t.add_mark(Mark::new("bold")); // -> true (added)
t.add_mark(Mark::new("bold")); // -> false (already present; deduped)
t.has_mark("bold"); // -> true
t.get_mark("bold"); // -> Option<&Mark>
t.toggle_mark(Mark::new("italic")); // add if absent, remove if present
t.set_mark_attr("link", "href", "https://tiptap.dev"); // set attr on an existing mark
t.remove_mark("bold"); // -> usize (count removed)
t.clear_marks(); // drop all marks
```
### Attributes
```rust
use tiptap_rusty_parser::Node;
use serde_json::json;
let mut h = Node::element("heading");
h.set_attr("level", 2); // -> previous value, if any
h.attr("level"); // -> Option<&Value> => Some(&json!(2))
h.attrs_mut().insert("class".into(), json!("title")); // raw map access
h.remove_attr("level"); // -> Option<Value>
```
### Children
```rust
use tiptap_rusty_parser::Node;
let mut p = Node::element("paragraph");
p.push_child(Node::text("a"));
p.push_child(Node::text("c"));
p.insert_child(1, Node::text("b")); // index clamped to len
p.child_count(); // -> 3
p.child(1); // -> Option<&Node>
p.child_mut(1); // -> Option<&mut Node>
p.children(); // -> &[Node]
p.children_mut(); // -> &mut Vec<Node> (creates if absent)
p.replace_child(0, Node::text("A")); // -> Option<Node> (old)
p.remove_child(2); // -> Option<Node> (removed)
p.retain_children(|c| c.get_text() != Some("A")); // filter in place
p.clear_children(); // remove all
```
### Text
```rust
# use tiptap_rusty_parser::Node;
let mut t = Node::text("old");
t.get_text(); // -> Some("old")
t.set_text("new");
```
### Bulk transforms
`replace_all` walks the whole subtree, applying a mutation to every node that
matches a predicate, and returns how many were changed.
```rust
use tiptap_rusty_parser::{Document, Mark};
let mut doc = Document::from_json_str(r#"{
"type":"doc","content":[
{"type":"paragraph","content":[{"type":"text","text":"x"}]},
{"type":"paragraph","content":[{"type":"text","text":"y"}]}
]}"#)?;
let changed = doc.replace_all(
|n| n.node_type.as_deref() == Some("text"),
|n| { n.add_mark(Mark::new("bold")); },
);
assert_eq!(changed, 2);
# Ok::<(), tiptap_rusty_parser::ParseError>(())
```
---
## Building nodes
Constructors plus consuming `with_*` builder methods for fluent assembly.
```rust
use tiptap_rusty_parser::{doc, Mark, Node};
// Leaf constructors
let plain = Node::text("hi");
let marked = Node::text_with_marks("bold!", [Mark::new("bold")]);
// Element builder
let para = Node::element("paragraph")
.with_attr("textAlign", "center")
.with_mark(Mark::new("bold"))
.with_text("hello") // adds a text child
.with_child(Node::text(" world")); // adds an arbitrary child
// Mark builder
let link = Mark::new("link").attr("href", "https://tiptap.dev");
// doc(..) helper for the root
let document = doc([
Node::element("heading").with_attr("level", 1).with_text("Title"),
para,
]);
```
| `Node::element(type)` | new element node of `type` |
| `Node::text(s)` | new `text` node |
| `Node::text_with_marks(s, marks)` | text node with marks |
| `doc(children)` | a `doc` root node |
| `Mark::new(type)` / `.attr(k, v)` | construct a mark |
| `.with_attr(k, v)` | set an attr (chaining) |
| `.with_child(node)` / `.with_children(iter)` | append child/children |
| `.with_text(s)` | append a text child |
| `.with_mark(mark)` | add a mark |
---
## Error handling
Parsing/serialization returns `Result<T, ParseError>`:
```rust
pub enum ParseError {
Json(serde_json::Error), // invalid JSON / shape mismatch
Io(std::io::Error), // reader failure (from_reader)
}
```
`ParseError` implements `std::error::Error` (via `thiserror`) and `From` for both
underlying errors, so `?` works directly. A `Result<T>` alias is also exported.
---
## Performance
Borrow-first API, stack-based descendant iteration (no recursion blowup on deep
docs), `serde_json` for (de)serialization, and a release profile with
`lto = true` / `codegen-units = 1`.
Indicative criterion baselines on a synthetic doc of **500 paragraphs × 20 bold
text spans** (~10k text nodes, ~10.5k nodes total):
| `parse` (from JSON string) | ~14 ms |
| `serialize` (to JSON string) | ~1.0 ms |
| `walk` (count all nodes) | ~29 µs |
| `find_all` (all text nodes) | ~108 µs |
| `replace_all` (add a mark to every text node) | ~5.0 ms |
Run `cargo bench` to reproduce on your hardware.
---
## Development
```sh
cargo test # unit + integration + doctests
cargo clippy --all-targets -- -D warnings
cargo bench # criterion baselines
```
---
## License
MIT