Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implement parsing of pushw and env io operations #40

Merged
merged 5 commits into from
Jan 7, 2022
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
223 changes: 203 additions & 20 deletions assembly/src/v1/parsers/io_ops.rs
Original file line number Diff line number Diff line change
Expand Up @@ -8,8 +8,48 @@ use super::{parse_element_param, AssemblyError, BaseElement, FieldElement, Opera
/// In cases when the immediate value is 0, PUSH operation is replaced with PAD. Also, in cases
/// when immediate value is 1, PUSH operation is replaced with PAD INCR because in most cases
/// this will be more efficient than doing a PUSH.
///
/// # Errors
///
/// This function expects an assembly op with exactly one immediate value that is a valid field
/// element in decimal or hexadecimal representation. It will return an error if the immediate
/// value is invalid or missing.
pub fn parse_push(span_ops: &mut Vec<Operation>, op: &Token) -> Result<(), AssemblyError> {
validate_op_len(op, 2, 2)?;

let value = parse_element_param(op, 1)?;
push_value(span_ops, value);

Ok(())
}

/// Appends PUSH operations to the span block until all elements of the provided word are pushed
/// onto the stack.
///
/// All immediate values are handled in the same way as for the single element "push" operation.
///
/// # Errors
///
/// This function expects an assembly op with 4 immediate values that are valid field elements
/// in decimal or hexadecimal representation. It will return an error if the assembly instruction's
/// immediate values are invalid.
pub fn parse_pushw(span_ops: &mut Vec<Operation>, op: &Token) -> Result<(), AssemblyError> {
validate_op_len(op, 5, 5)?;

for idx in 1..=4 {
let value = parse_element_param(op, idx)?;
push_value(span_ops, value);
}

Ok(())
}

/// This is a helper function that appends a PUSH operation to the span block which puts the
/// provided value parameter onto the stack.
///
/// When the value is 0, PUSH operation is replaced with PAD. When the value is 1, PUSH operation
/// is replaced with PAD INCR because in most cases this will be more efficient than doing a PUSH.
fn push_value(span_ops: &mut Vec<Operation>, value: BaseElement) {
if value == BaseElement::ZERO {
span_ops.push(Operation::Pad);
} else if value == BaseElement::ONE {
Expand All @@ -18,20 +58,33 @@ pub fn parse_push(span_ops: &mut Vec<Operation>, op: &Token) -> Result<(), Assem
} else {
span_ops.push(Operation::Push(value));
}
Ok(())
}

/// TODO: implement
pub fn parse_pushw(_span_ops: &mut Vec<Operation>, _op: &Token) -> Result<(), AssemblyError> {
unimplemented!()
}

// ENVIRONMENT INPUTS
// ================================================================================================

/// TODO: implement
pub fn parse_env(_span_ops: &mut Vec<Operation>, _op: &Token) -> Result<(), AssemblyError> {
unimplemented!()
/// Appends machine operations to the current span block according to the requested environment
/// assembly instruction.
///
/// "env.sdepth" pushes the current depth of the stack onto the top of the stack, which is handled
/// directly by the SDEPTH operation.
///
/// # Errors
///
/// This function expects a valid assembly environment op that specifies the environment input to
/// be handled. It will return an error if the assembly instruction is malformed or the environment
/// input is unrecognized.
pub fn parse_env(span_ops: &mut Vec<Operation>, op: &Token) -> Result<(), AssemblyError> {
validate_op_len(op, 2, 2)?;

match op.parts()[1] {
"sdepth" => {
span_ops.push(Operation::SDepth);
}
_ => return Err(AssemblyError::invalid_op(op)),
}

Ok(())
}

// NON-DETERMINISTIC INPUTS
Expand Down Expand Up @@ -64,14 +117,12 @@ pub fn parse_readw(_span_ops: &mut Vec<Operation>, _op: &Token) -> Result<(), As
/// "mem.store" is a write operation that saves the top 4 elements of the stack to memory and
/// leaves them on the stack.
pub fn parse_mem(span_ops: &mut Vec<Operation>, op: &Token) -> Result<(), AssemblyError> {
match op.num_parts() {
0 | 1 => Err(AssemblyError::missing_param(op)),
2 | 3 => match op.parts()[1] {
"push" | "load" => parse_mem_read(span_ops, op),
"pop" | "store" => parse_mem_write(span_ops, op),
_ => Err(AssemblyError::invalid_op(op)),
},
_ => Err(AssemblyError::extra_param(op)),
validate_op_len(op, 2, 3)?;

match op.parts()[1] {
"push" | "load" => parse_mem_read(span_ops, op),
"pop" | "store" => parse_mem_write(span_ops, op),
_ => Err(AssemblyError::invalid_op(op)),
}
}

Expand Down Expand Up @@ -151,6 +202,26 @@ fn push_mem_addr(span_ops: &mut Vec<Operation>, op: &Token) -> Result<(), Assemb
Ok(())
}

// HELPERS
// ================================================================================================

/// This is a helper function that validates the length of the assembly io instruction and returns
/// an error if params are missing or there are extras.
///
/// The minimum and maximum number of valid instruction parts for the op are provided by the
/// caller.
///
/// # Errors
///
/// This function will return an AssemblyError if the assembly op has too many or too few parts.
fn validate_op_len(op: &Token, min_parts: usize, max_parts: usize) -> Result<(), AssemblyError> {
match op.num_parts() {
too_few if too_few < min_parts => Err(AssemblyError::missing_param(op)),
too_many if too_many > max_parts => Err(AssemblyError::extra_param(op)),
_ => Ok(()),
}
}

Comment on lines +208 to +224
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not for this PR, but I wonder if we should move this out to be used in other modules as well.

Copy link
Contributor Author

@grjte grjte Jan 9, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I think we should. Otherwise we're either duplicating code or we're under-validating. Pulling this logic out & calling the function instead of using match should make all of the op parsing implementations more readable.

// TESTS
// ================================================================================================

Expand All @@ -163,21 +234,114 @@ mod tests {
let mut span_ops: Vec<Operation> = Vec::new();
let op_0 = Token::new("push.0", 0);
let op_1 = Token::new("push.1", 0);
let op_other = Token::new("push.2", 0);
let op_dec = Token::new("push.135", 0);
let op_hex = Token::new("push.0x7b", 0);
let expected = vec![
Operation::Pad,
Operation::Pad,
Operation::Incr,
Operation::Push(BaseElement::new(2)),
Operation::Push(BaseElement::new(135)),
Operation::Push(BaseElement::new(123)),
];

parse_push(&mut span_ops, &op_0).expect("Failed to parse push.0");
parse_push(&mut span_ops, &op_1).expect("Failed to parse push.1");
parse_push(&mut span_ops, &op_other).expect("Failed to parse push.2");
parse_push(&mut span_ops, &op_dec).expect("Failed to parse push of decimal element 123");
parse_push(&mut span_ops, &op_hex).expect("Failed to parse push of hex element 0x7b");

assert_eq!(span_ops, expected);
}

#[test]
fn push_invalid() {
// fails when immediate value is invalid or missing
let mut span_ops: Vec<Operation> = Vec::new();
let param_idx = 0;

// value missing
let op_no_val = Token::new("pushw", param_idx);
assert!(parse_push(&mut span_ops, &op_no_val).is_err());

// invalid value
let op_val_invalid = Token::new("push.abc", param_idx);
assert!(parse_push(&mut span_ops, &op_val_invalid).is_err());

// extra value
let op_extra_val = Token::new("pushw.0.1", param_idx);
assert!(parse_push(&mut span_ops, &op_extra_val).is_err());
}

#[test]
fn pushw() {
// pushes a word of 4 immediate values in decimal or hexadecimal onto the stack
let mut span_ops: Vec<Operation> = Vec::new();
let op = Token::new("pushw.1.23.0x1C8.0", 0);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not for this PR: now that I'm seeing how this actually looks, I wonder if we should change the format of pushw instruction to always take a single hexadecimal value of 32 bytes. Or maybe making this as an additional allowed format. So, for example, possible formats could be:

  • pushw.1.2.3.4
  • pushw.0x13244... (32 bytes)

Let's create an issue to think this through later on.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I opened an issue here: #44

let expected = vec![
Operation::Pad,
Operation::Incr,
Operation::Push(BaseElement::new(23)),
Operation::Push(BaseElement::new(456)),
Operation::Pad,
];
parse_pushw(&mut span_ops, &op).expect("Failed to parse pushw");

assert_eq!(span_ops, expected);
}

#[test]
fn pushw_invalid() {
// fails when immediate values are invalid or missing
let mut span_ops: Vec<Operation> = Vec::new();
let param_idx = 0;

// no values
let op_no_vals = Token::new("pushw", param_idx);
assert!(parse_pushw(&mut span_ops, &op_no_vals).is_err());

// insufficient values provided
let op_val_missing = Token::new("pushw.0.1.2", param_idx);
assert!(parse_pushw(&mut span_ops, &op_val_missing).is_err());

// invalid value
let op_val_invalid = Token::new("push.0.1.2.abc", param_idx);
assert!(parse_pushw(&mut span_ops, &op_val_invalid).is_err());

// extra value
let op_extra_val = Token::new("pushw.0.1.2.3.4", param_idx);
assert!(parse_pushw(&mut span_ops, &op_extra_val).is_err());
}

#[test]
fn env_sdepth() {
// pushes the current depth of the stack onto the top of the stack
let mut span_ops = vec![Operation::Push(BaseElement::ONE); 8];
let op = Token::new("env.sdepth", 0);
let mut expected = span_ops.clone();
expected.push(Operation::SDepth);

parse_env(&mut span_ops, &op).expect("Failed to parse env.sdepth with empty stack");
assert_eq!(span_ops, expected);
}

#[test]
fn env_invalid() {
// fails when env op variant is invalid or missing or has too many immediate values
let mut span_ops: Vec<Operation> = Vec::new();
let param_idx = 0;

// missing env var
let op_no_val = Token::new("env", param_idx);
assert!(parse_mem(&mut span_ops, &op_no_val).is_err());

// invalid env var
let op_val_invalid = Token::new("env.invalid", param_idx);
assert!(parse_push(&mut span_ops, &op_val_invalid).is_err());

// extra value
let op_extra_val = Token::new("env.sdepth.0", param_idx);
assert!(parse_push(&mut span_ops, &op_extra_val).is_err());
}

#[test]
fn mem_pop() {
// stores the top 4 elements of the stack in memory
Expand Down Expand Up @@ -298,4 +462,23 @@ mod tests {

assert_eq!(&span_ops_addr, &expected_addr);
}

#[test]
fn mem_invalid() {
// fails when mem op variant is invalid or missing or has too many immediate values
let mut span_ops: Vec<Operation> = Vec::new();
let param_idx = 0;

// missing variant
let op_no_val = Token::new("mem", param_idx);
assert!(parse_mem(&mut span_ops, &op_no_val).is_err());

// invalid variant
let op_val_invalid = Token::new("mem.abc", param_idx);
assert!(parse_push(&mut span_ops, &op_val_invalid).is_err());

// extra value
let op_extra_val = Token::new("mem.push.0.1", param_idx);
assert!(parse_push(&mut span_ops, &op_extra_val).is_err());
}
}