Add serialization support (#171)

* Add serialization

* Add tests and fix a few bugs

* fmt

* Add documentation for using insta

* Try adding all snapshot files

Asked by @oflatt
https://egraphs.zulipchat.com/#narrow/stream/328972-general/topic/e-graph.20serialization.20format/near/376503331

* Remove snapshot tests for now

Cargo.toml

@@ -13,7 +13,7 @@ name = "files"
 [features]
 default = ["bin"]
 
-bin = ["dep:clap", "dep:env_logger"]
+bin = ["dep:clap", "dep:env_logger", "egraph-serialize/serde", "dep:serde_json"]
 wasm-bindgen = ["instant/wasm-bindgen"]
 
 [dependencies]
@@ -33,6 +33,9 @@ num-traits = "0.2.15"
 smallvec = "1.11"
 symbolic_expressions = {git = "https://github.com/oflatt/symbolic-expressions", rev = "4c0ea5ca008f972450b2af72387e64d2c1c6a791"}
 
+egraph-serialize = {git = "https://github.com/egraphs-good/egraph-serialize", rev = "54b1a4f1e2f2135846b084edcb495cd159839540", default-features = false}
+serde_json ={optional=true, version = "1.0.100", features = ["preserve_order"]}
+
 lalrpop-util = {version = "0.20", features = ["lexer"]}
 regex = "1"
 

README.md

@@ -9,7 +9,7 @@ This is repo for the `egglog` tool accompanying the paper
   "Better Together: Unifying Datalog and Equality Saturation"
   ([ACM DL](https://dl.acm.org/doi/10.1145/3591239), [arXiv](https://arxiv.org/abs/2304.04332)).
 
-If you use this work, please use [this citation](./CITATION.bib). 
+If you use this work, please use [this citation](./CITATION.bib).
 
 See also the Python binding, which provides a bit more documentation:
 https://egg-smol-python.readthedocs.io/en/latest/
@@ -30,7 +30,7 @@ make all
 ## Usage
 
 ```
-cargo run [-f fact-path] [-naive] <files.egg>
+cargo run [-f fact-path] [-naive] [--to-json] <files.egg>
 ```
 
 or just
@@ -63,6 +63,10 @@ npm install -g @vscode/vsce
 
 Run `vsce package` in the `vscode/egglog-1.0.0` folder to reconstruct the .vsix file and install it manually.
 
+## Development
+
+To run the tests use `make test`.
+
 # Syntax
 
 The syntax of the .egg files is defined in `src/ast/parse.lalrpop`.
@@ -138,7 +142,7 @@ If you define a `:merge` expression, you can update specific values in the funct
 (set (KeepMax 1) 1)
 (set (KeepMax 1) 2)   ; we redefine 1 to be 2
 (set (KeepMax 1) 0)   ; this does not change since we use max
-(extract (KeepMax 1)) ; this is 2  
+(extract (KeepMax 1)) ; this is 2
 ```
 
 ### `relation` command
@@ -240,7 +244,7 @@ Example:
          (Add b a))
 ```
 
-declares a rule that a `Add` variant is commutative. 
+declares a rule that a `Add` variant is commutative.
 
 ```
 (birewrite (* (* a b) c) (* a (* b c)))
@@ -326,7 +330,7 @@ Union only works on variants, not sorts.
 ### Name
 
 ```
-    [ <Ident> ] 
+    [ <Ident> ]
 ```
 
 ### Facts

src/function/mod.rs

@@ -4,7 +4,7 @@ use smallvec::SmallVec;
 
 mod binary_search;
 pub mod index;
-mod table;
+pub(crate) mod table;
 
 pub type ValueVec = SmallVec<[Value; 3]>;
 

src/function/table.rs

@@ -51,7 +51,7 @@ pub(crate) struct Table {
     max_ts: u32,
     n_stale: usize,
     table: RawTable<TableOffset>,
-    vals: Vec<(Input, TupleOutput)>,
+    pub(crate) vals: Vec<(Input, TupleOutput)>,
 }
 
 /// Used for the RawTable probe sequence.
@@ -299,7 +299,7 @@ impl Table {
     }
 }
 
-fn hash_values(vs: &[Value]) -> u64 {
+pub(crate) fn hash_values(vs: &[Value]) -> u64 {
     // Just hash the bits: all inputs to the same function should have matching
     // column types.
     let mut hasher = BH::default().build_hasher();
@@ -310,8 +310,8 @@ fn hash_values(vs: &[Value]) -> u64 {
 }
 
 #[derive(Debug, Clone, PartialEq, Eq)]
-struct Input {
-    data: ValueVec,
+pub(crate) struct Input {
+    pub(crate) data: ValueVec,
     /// The timestamp at which the given input became "stale"
     stale_at: u32,
 }
@@ -328,7 +328,7 @@ impl Input {
         self.data.as_slice()
     }
 
-    fn live(&self) -> bool {
+    pub(crate) fn live(&self) -> bool {
         self.stale_at == u32::MAX
     }
 }

src/lib.rs

@@ -3,6 +3,7 @@ mod extract;
 mod function;
 mod gj;
 mod proofs;
+mod serialize;
 pub mod sort;
 mod typecheck;
 mod typechecking;
@@ -13,6 +14,7 @@ mod value;
 use hashbrown::hash_map::Entry;
 use index::ColumnIndex;
 use instant::{Duration, Instant};
+pub use serialize::SerializeConfig;
 use sort::*;
 use thiserror::Error;
 

src/main.rs

@@ -1,5 +1,5 @@
 use clap::Parser;
-use egglog::EGraph;
+use egglog::{EGraph, SerializeConfig};
 use std::io::{self, BufRead, BufReader};
 use std::path::PathBuf;
 
@@ -12,6 +12,8 @@ struct Args {
     inputs: Vec<PathBuf>,
     #[clap(long)]
     proofs: bool,
+    #[clap(long)]
+    to_json: bool,
 }
 
 fn main() {
@@ -75,6 +77,12 @@ fn main() {
             }
         }
 
+        if args.to_json {
+            let json_path = input.with_extension("json");
+            let serialized = egraph.serialize(SerializeConfig::default());
+            serialized.to_json_file(json_path).unwrap();
+        }
+
         // no need to drop the egraph if we are going to exit
         if idx == args.inputs.len() - 1 {
             std::mem::forget(egraph)