UPDATE: Implement the implicit Intersection Operator

The II operator takes a range and returns a single cell that is in the same column or the same row
as the present cell.

This is needed for backwards compatibility with old Excel models and as a first step towards dynamic arrays.

In the past Excel would evaluate `=A1:A10` in cell `C3` as `A3`, but today in results in an array containing all
values in the range. To be compatible with old workbooks Excel inserts the II operator
on those cases.

So this PR performs an static analysis on all formulas inserting on import automatically the II operator
where necessary. This we call the _automatic implicit operator_. When exporting to Excel the operator is striped away.
You can also manually use the II. For instance `=SUM(@A1:A10)` in cell `C3`.
This was not possible before and such a formula would break backwards compatibility with Excel. To Excel that "non automatic"
form of the II is exported as `_xlfn.SINGLE()`.

Th static analysis has to be done for all arithmetic operations and all functions.
This is a bit of a daunting task and it is not done fully in this PR. We also need to implement arrays and dynamic arrays.
My believe is that once the core operations have been implemented we can go formula by formula writing proper tests and documentation.

After this PR formulas like `=A1:A10` for instance will return `#N/IMPL!` instead of performing the implicit intersection
This commit is contained in:
Nicolás Hatcher
2024-11-20 00:35:19 +01:00
committed by Nicolás Hatcher Andrés
parent 90763048bc
commit da017b6113
31 changed files with 1623 additions and 694 deletions

View File

@@ -1,5 +1,5 @@
/*!
# GRAMAR
# GRAMMAR
<pre class="rust">
opComp => '=' | '<' | '>' | '<=' } '>=' | '<>'
@@ -12,7 +12,8 @@ term => factor (opFactor factor)*
factor => prod (opProd prod)*
prod => power ('^' power)*
power => (unaryOp)* range '%'*
range => primary (':' primary)?
range => implicit (':' primary)?
implicit=> '@' primary | primary
primary => '(' expr ')'
=> number
=> function '(' f_args ')'
@@ -45,8 +46,8 @@ use super::utils::number_to_column;
use token::OpCompare;
pub mod move_formula;
pub mod static_analysis;
pub mod stringify;
pub mod walk;
#[cfg(test)]
mod tests;
@@ -81,6 +82,9 @@ fn get_table_column_by_name(table_column_name: &str, table: &Table) -> Option<i3
None
}
// DefinedNameS is a tuple with the name of the defined name, the index of the sheet and the formula
pub type DefinedNameS = (String, Option<u32>, String);
pub(crate) struct Reference<'a> {
sheet_name: &'a Option<String>,
sheet_index: u32,
@@ -164,9 +168,13 @@ pub enum Node {
args: Vec<Node>,
},
ArrayKind(Vec<Node>),
DefinedNameKind((String, Option<u32>)),
DefinedNameKind(DefinedNameS),
TableNameKind(String),
WrongVariableKind(String),
ImplicitIntersection {
automatic: bool,
child: Box<Node>,
},
CompareKind {
kind: OpCompare,
left: Box<Node>,
@@ -189,7 +197,7 @@ pub enum Node {
pub struct Parser {
lexer: lexer::Lexer,
worksheets: Vec<String>,
defined_names: Vec<(String, Option<u32>)>,
defined_names: Vec<DefinedNameS>,
context: CellReferenceRC,
tables: HashMap<String, Table>,
}
@@ -197,7 +205,7 @@ pub struct Parser {
impl Parser {
pub fn new(
worksheets: Vec<String>,
defined_names: Vec<(String, Option<u32>)>,
defined_names: Vec<DefinedNameS>,
tables: HashMap<String, Table>,
) -> Parser {
let lexer = lexer::Lexer::new(
@@ -228,7 +236,7 @@ impl Parser {
pub fn set_worksheets_and_names(
&mut self,
worksheets: Vec<String>,
defined_names: Vec<(String, Option<u32>)>,
defined_names: Vec<DefinedNameS>,
) {
self.worksheets = worksheets;
self.defined_names = defined_names;
@@ -252,17 +260,17 @@ impl Parser {
// Returns:
// * None: If there is no defined name by that name
// * Some(Some(index)): If there is a defined name local to that sheet
// * Some((Some(index), formula)): If there is a defined name local to that sheet
// * Some(None): If there is a global defined name
fn get_defined_name(&self, name: &str, sheet: u32) -> Option<Option<u32>> {
for (df_name, df_scope) in &self.defined_names {
fn get_defined_name(&self, name: &str, sheet: u32) -> Option<(Option<u32>, String)> {
for (df_name, df_scope, df_formula) in &self.defined_names {
if name.to_lowercase() == df_name.to_lowercase() && df_scope == &Some(sheet) {
return Some(*df_scope);
return Some((*df_scope, df_formula.to_owned()));
}
}
for (df_name, df_scope) in &self.defined_names {
for (df_name, df_scope, df_formula) in &self.defined_names {
if name.to_lowercase() == df_name.to_lowercase() && df_scope.is_none() {
return Some(None);
return Some((None, df_formula.to_owned()));
}
}
None
@@ -411,7 +419,7 @@ impl Parser {
}
fn parse_range(&mut self) -> Node {
let t = self.parse_primary();
let t = self.parse_implicit();
if let Node::ParseErrorKind { .. } = t {
return t;
}
@@ -430,6 +438,22 @@ impl Parser {
t
}
fn parse_implicit(&mut self) -> Node {
let next_token = self.lexer.peek_token();
if next_token == TokenType::At {
self.lexer.advance_token();
let t = self.parse_primary();
if let Node::ParseErrorKind { .. } = t {
return t;
}
return Node::ImplicitIntersection {
automatic: false,
child: Box::new(t),
};
}
self.parse_primary()
}
fn parse_primary(&mut self) -> Node {
let next_token = self.lexer.next_token();
match next_token {
@@ -604,6 +628,20 @@ impl Parser {
args,
};
}
if &name == "_xlfn.SINGLE" {
if args.len() != 1 {
return Node::ParseErrorKind {
formula: self.lexer.get_formula(),
position: self.lexer.get_position() as usize,
message: "Implicit Intersection requires just one argument"
.to_string(),
};
}
return Node::ImplicitIntersection {
automatic: false,
child: Box::new(args[0].clone()),
};
}
return Node::InvalidFunctionKind { name, args };
}
let context = &self.context;
@@ -620,8 +658,8 @@ impl Parser {
};
// Could be a defined name or a table
if let Some(scope) = self.get_defined_name(&name, context_sheet_index) {
return Node::DefinedNameKind((name, scope));
if let Some((scope, formula)) = self.get_defined_name(&name, context_sheet_index) {
return Node::DefinedNameKind((name, scope, formula));
}
let name_lower = name.to_lowercase();
for table_name in self.tables.keys() {
@@ -706,6 +744,14 @@ impl Parser {
message: "Unexpected token: 'POWER'".to_string(),
}
}
TokenType::At => {
// A primary Node cannot start with an operator
Node::ParseErrorKind {
formula: self.lexer.get_formula(),
position: 0,
message: "Unexpected token: '@'".to_string(),
}
}
TokenType::RightParenthesis
| TokenType::RightBracket
| TokenType::Colon