Files
IronCalc/base/src/arithmetic.rs
Nicolás Hatcher e5ec75495a UPDATE: Introducing Arrays
# This PR introduces:

## Parsing arrays:

{1,2,3} and {1;2;3}

Note that array elements can be numbers, booleans and errors (#VALUE!)

## Evaluating arrays in the SUM function

=SUM({1,2,3}) works!

## Evaluating arithmetic operation with arrays

=SUM({1,2,3} * 8) or =SUM({1,2,3}+{2,4,5}) works

This is done with just one function (handle_arithmetic) for most operations

## Some mathematical functions implement arrays

=SUM(SIN({1,2,3})) works

This is done with macros. See fn_single_number
So that implementing new functions that supports array are easy


# Not done in this PR

## Most functions are not supporting arrays

When that happens we either through #N/IMPL! (not implemented error)
or do implicit intersection. Some functions will be rather trivial to "arraify" some will be hard

## The final result in a cell cannot be an array

The formula ={1,2,3} in a cell will result in #N/IMPL!

## Exporting arrays to Excel might not work correctly

Excel uses the cm (cell metadata) for formulas that contain dynamic arrays.
Although the present PR does not introduce dynamic arrays some formulas like =SUM(SIN({1,2,3}))
is considered a dynamic formula

## There are not a lot of tests in this delivery

The bulk of the tests will be added once we start going function by function# This PR introduces:

## Parsing arrays:

{1,2,3} and {1;2;3}

Note that array elements can be numbers, booleans and errors (#VALUE!)

## Evaluating arrays in the SUM function

=SUM({1,2,3}) works!

## Evaluating arithmetic operation with arrays

=SUM({1,2,3} * 8) or =SUM({1,2,3}+{2,4,5}) works

This is done with just one function (handle_arithmetic) for most operations

## Some mathematical functions implement arrays

=SUM(SIN({1,2,3})) works

This is done with macros. See fn_single_number
So that implementing new functions that supports array are easy


# Not done in this PR

## Most functions are not supporting arrays

When that happens we either through #N/IMPL! (not implemented error)
or do implicit intersection. Some functions will be rather trivial to "arraify" some will be hard

## The final result in a cell cannot be an array

The formula ={1,2,3} in a cell will result in #N/IMPL!

## Exporting arrays to Excel might not work correctly

Excel uses the cm (cell metadata) for formulas that contain dynamic arrays.
Although the present PR does not introduce dynamic arrays some formulas like =SUM(SIN({1,2,3}))
is considered a dynamic formula

## There are not a lot of tests in this delivery

The bulk of the tests will be added once we start going function by function

## The array parsing does not respect the locale

Locales that use ',' as a decimal separator need to use something different for arrays

## The might introduce a small performance penalty

We haven't been benchmarking, and having closures for every arithmetic operation and every function
evaluation will introduce a performance hit. Fixing that in he future is not so hard writing tailored
code for the operation
2025-03-17 20:04:47 +01:00

159 lines
6.4 KiB
Rust
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

use crate::{
calc_result::CalcResult,
cast::NumberOrArray,
expressions::{
parser::{ArrayNode, Node},
token::Error,
types::CellReferenceIndex,
},
model::Model,
};
/// Unify how we map booleans/strings to f64
fn to_f64(value: &ArrayNode) -> Result<f64, Error> {
match value {
ArrayNode::Number(f) => Ok(*f),
ArrayNode::Boolean(b) => Ok(if *b { 1.0 } else { 0.0 }),
ArrayNode::String(s) => match s.parse::<f64>() {
Ok(f) => Ok(f),
Err(_) => Err(Error::VALUE),
},
ArrayNode::Error(err) => Err(err.clone()),
}
}
impl Model {
/// Applies `op` elementwise for arrays/numbers.
pub(crate) fn handle_arithmetic(
&mut self,
left: &Node,
right: &Node,
cell: CellReferenceIndex,
op: &dyn Fn(f64, f64) -> Result<f64, Error>,
) -> CalcResult {
let l = match self.get_number_or_array(left, cell) {
Ok(f) => f,
Err(s) => {
return s;
}
};
let r = match self.get_number_or_array(right, cell) {
Ok(f) => f,
Err(s) => {
return s;
}
};
match (l, r) {
// -----------------------------------------------------
// Case 1: Both are numbers
// -----------------------------------------------------
(NumberOrArray::Number(f1), NumberOrArray::Number(f2)) => match op(f1, f2) {
Ok(x) => CalcResult::Number(x),
Err(Error::DIV) => CalcResult::Error {
error: Error::DIV,
origin: cell,
message: "Divide by 0".to_string(),
},
Err(Error::VALUE) => CalcResult::Error {
error: Error::VALUE,
origin: cell,
message: "Invalid number".to_string(),
},
Err(e) => CalcResult::Error {
error: e,
origin: cell,
message: "Unknown error".to_string(),
},
},
// -----------------------------------------------------
// Case 2: left is Number, right is Array
// -----------------------------------------------------
(NumberOrArray::Number(f1), NumberOrArray::Array(a2)) => {
let mut array = Vec::new();
for row in a2 {
let mut data_row = Vec::new();
for node in row {
match to_f64(&node) {
Ok(f2) => match op(f1, f2) {
Ok(x) => data_row.push(ArrayNode::Number(x)),
Err(Error::DIV) => data_row.push(ArrayNode::Error(Error::DIV)),
Err(Error::VALUE) => data_row.push(ArrayNode::Error(Error::VALUE)),
Err(e) => data_row.push(ArrayNode::Error(e)),
},
Err(err) => data_row.push(ArrayNode::Error(err)),
}
}
array.push(data_row);
}
CalcResult::Array(array)
}
// -----------------------------------------------------
// Case 3: left is Array, right is Number
// -----------------------------------------------------
(NumberOrArray::Array(a1), NumberOrArray::Number(f2)) => {
let mut array = Vec::new();
for row in a1 {
let mut data_row = Vec::new();
for node in row {
match to_f64(&node) {
Ok(f1) => match op(f1, f2) {
Ok(x) => data_row.push(ArrayNode::Number(x)),
Err(Error::DIV) => data_row.push(ArrayNode::Error(Error::DIV)),
Err(Error::VALUE) => data_row.push(ArrayNode::Error(Error::VALUE)),
Err(e) => data_row.push(ArrayNode::Error(e)),
},
Err(err) => data_row.push(ArrayNode::Error(err)),
}
}
array.push(data_row);
}
CalcResult::Array(array)
}
// -----------------------------------------------------
// Case 4: Both are arrays
// -----------------------------------------------------
(NumberOrArray::Array(a1), NumberOrArray::Array(a2)) => {
let n1 = a1.len();
let m1 = a1.first().map(|r| r.len()).unwrap_or(0);
let n2 = a2.len();
let m2 = a2.first().map(|r| r.len()).unwrap_or(0);
let n = n1.max(n2);
let m = m1.max(m2);
let mut array = Vec::new();
for i in 0..n {
let row1 = a1.get(i);
let row2 = a2.get(i);
let mut data_row = Vec::new();
for j in 0..m {
let val1 = row1.and_then(|r| r.get(j));
let val2 = row2.and_then(|r| r.get(j));
match (val1, val2) {
(Some(v1), Some(v2)) => match (to_f64(v1), to_f64(v2)) {
(Ok(f1), Ok(f2)) => match op(f1, f2) {
Ok(x) => data_row.push(ArrayNode::Number(x)),
Err(Error::DIV) => data_row.push(ArrayNode::Error(Error::DIV)),
Err(Error::VALUE) => {
data_row.push(ArrayNode::Error(Error::VALUE))
}
Err(e) => data_row.push(ArrayNode::Error(e)),
},
(Err(e), _) | (_, Err(e)) => data_row.push(ArrayNode::Error(e)),
},
// Mismatched dimensions => #VALUE!
_ => data_row.push(ArrayNode::Error(Error::VALUE)),
}
}
array.push(data_row);
}
CalcResult::Array(array)
}
}
}
}