Validation now outputs a Validated Interpreter Operation

This commit is contained in:
Yuriy Dupyn 2024-01-27 21:22:00 +01:00
parent 935d9814ae
commit 562e732138
6 changed files with 133 additions and 120 deletions

View file

@ -1,4 +1,5 @@
use minisql::{operation::Operation, schema::TableSchema};
use crate::syntax::RawQuerySyntax;
use nom::{branch::alt, multi::many0, IResult};
use thiserror::Error;
@ -34,7 +35,9 @@ pub fn parse_and_validate(query: String, db_metadata: &Vec<(String, &TableSchema
Error::ParsingError(err.to_string())
})?;
validate_operation(&op, db_metadata)?;
// TODO
// validate_operation(&op, db_metadata)?;
todo!();
Ok(op)
}

View file

@ -2,6 +2,7 @@
mod parsing;
mod validation;
mod core;
mod syntax;
pub use core::parse_and_validate;
pub use core::Error;

36
parser/src/syntax.rs Normal file
View file

@ -0,0 +1,36 @@
use minisql::{type_system::Value, schema::{TableSchema, ColumnName, TableName}};
// TODO: Move this out into separate file and rename to something like Syntax, SyntaxTree,
// OperationSyntax, RawOperationSyntax
pub enum RawQuerySyntax {
Select(TableName, ColumnSelection, Option<Condition>),
Insert(TableName, InsertionValues),
Delete(TableName, Option<Condition>),
// Update(...),
CreateTable(TableName, TableSchema),
CreateIndex(TableName, ColumnName),
// DropTable(TableName),
}
pub type InsertionValues = Vec<(ColumnName, Value)>;
pub enum ColumnSelection {
All,
Columns(Vec<ColumnName>),
}
pub enum Condition {
// And(Box<Condition>, Box<Condition>),
// Or(Box<Condition>, Box<Condition>),
// Not(Box<Condition>),
Eq(ColumnName, Value),
// LessOrEqual(ColumnName, DbValue),
// Less(ColumnName, DbValue),
// StringCondition(StringCondition),
}
// enum StringCondition {
// Prefix(ColumnName, String),
// Substring(ColumnName, String),
// }

View file

@ -1,9 +1,9 @@
use std::collections::HashSet;
use std::collections::HashMap;
use thiserror::Error;
use minisql::{operation::{ColumnSelection, Condition, InsertionValues, Operation}, schema::{TableSchema, ColumnName, TableName}, type_system::DbType};
use crate::syntax::{ColumnSelection, Condition, InsertionValues, RawQuerySyntax};
use minisql::{operation::{ColumnSelectionForInterpreter, ConditionForInterpreter, InsertionValuesForInterpreter, OperationForInterpreter}, type_system::Value, schema::{TableSchema, ColumnName, TableName}, type_system::DbType, interpreter::TablePosition};
#[derive(Debug, Error)]
pub enum ValidationError {
@ -25,69 +25,55 @@ pub enum ValidationError {
RequiredColumnsAreMissing(Vec<ColumnName>)
}
pub type DbSchema<'a> = Vec<(TableName, &'a TableSchema)>;
pub type DbSchema<'a> = Vec<(TableName, TablePosition, &'a TableSchema)>;
/// Validates the operation based on db_metadata
pub fn validate_operation(operation: &Operation, db_schema: &DbSchema) -> Result<(), ValidationError> {
match operation {
Operation::Select(table_name, column_selection, condition) => {
validate_select(table_name, column_selection, condition, db_schema)?;
/// Validates and converts the raw syntax into a proper interpreter operation based on db schema.
pub fn validate_operation(query: RawQuerySyntax, db_schema: &DbSchema) -> Result<OperationForInterpreter, ValidationError> {
match query {
RawQuerySyntax::Select(table_name, column_selection, condition) => {
validate_select(table_name, column_selection, condition, db_schema)
},
Operation::Insert(table_name, insertion_values) => {
validate_insert(&table_name, insertion_values, db_schema)?;
RawQuerySyntax::Insert(table_name, insertion_values) => {
validate_insert(table_name, insertion_values, db_schema)
},
Operation::Delete(table_name, condition) => {
validate_delete(table_name, condition, db_schema)?;
RawQuerySyntax::Delete(table_name, condition) => {
validate_delete(table_name, condition, db_schema)
},
// Operation::Update(table_name, insertion_values, condition) => {
// validate_update(table_name, insertion_values, db_metadata)?;
// },
Operation::CreateTable(table_name, schema) => {
validate_create(table_name, schema, db_schema)?;
RawQuerySyntax::CreateTable(table_name, schema) => {
validate_create(table_name, schema, db_schema)
},
Operation::CreateIndex(table_name, column_name) => {
validate_create_index(table_name, column_name, db_schema)?;
RawQuerySyntax::CreateIndex(table_name, column_name) => {
validate_create_index(table_name, column_name, db_schema)
},
// Operation::DropTable(table_name) => {
// validate_drop(table_name, db_schema)?;
// }
}
Ok(())
}
fn validate_table_exists<'a>(db_schema: &DbSchema<'a>, table_name: &'a TableName) -> Result<&'a TableSchema, ValidationError> {
db_schema.iter().find(|(tname, _)| table_name.eq(tname))
fn validate_table_exists<'a>(db_schema: &DbSchema<'a>, table_name: &'a TableName) -> Result<(TablePosition, &'a TableSchema), ValidationError> {
db_schema.iter().find(|(tname, _, _)| table_name.eq(tname))
.ok_or(ValidationError::TableDoesNotExist(table_name.to_string()))
.map(|(_, table_schema)| table_schema).copied()
.map(|(_, table_position, table_schema)| (*table_position, *table_schema))
}
// pub fn validate_drop(table_name: &str, db_metadata: &Vec<(String, TableSchema)>) -> Result<(), ValidationError> {
// db_metadata.iter().find(|(tname, _)| table_name.eq(tname))
// .ok_or(ValidationError::TableDoesNotExist(table_name.to_string()))?;
// Ok(())
// }
pub fn validate_create(table_name: &TableName, schema: &TableSchema, db_schema: &DbSchema) -> Result<(), ValidationError> {
if let Some(_) = get_table_schema(db_schema, table_name) {
pub fn validate_create(table_name: TableName, table_schema: TableSchema, db_schema: &DbSchema) -> Result<OperationForInterpreter, ValidationError> {
if let Some(_) = get_table_schema(db_schema, &table_name) {
return Err(ValidationError::TableAlreadyExists(table_name.to_string()));
}
find_first_duplicate(&schema.get_columns())
find_first_duplicate(&table_schema.get_columns())
.map_or_else(
|| Ok(()),
|duplicate_column| Err(ValidationError::DuplicateColumn(duplicate_column.to_string()))
)?;
// TODO: Ensure it has a primary key??
Ok(())
Ok(OperationForInterpreter::CreateTable(table_name, table_schema))
}
pub fn validate_select(table_name: &TableName, column_selection: &ColumnSelection, condition: &Option<Condition>, db_schema: &Vec<(TableName, &TableSchema)>) -> Result<(), ValidationError> {
let schema = validate_table_exists(db_schema, table_name)?;
pub fn validate_select(table_name: TableName, column_selection: ColumnSelection, condition: Option<Condition>, db_schema: &DbSchema) -> Result<OperationForInterpreter, ValidationError> {
let (table_position, schema) = validate_table_exists(db_schema, &table_name)?;
match column_selection {
ColumnSelection::Columns(columns) => {
let non_existant_columns: Vec<String> =
let non_existant_columns: Vec<ColumnName> =
columns.iter().filter_map(|column|
if schema.does_column_exist(&column) {
Some(column.clone())
@ -97,52 +83,21 @@ pub fn validate_select(table_name: &TableName, column_selection: &ColumnSelectio
if non_existant_columns.len() > 0 {
Err(ValidationError::ColumnsDoNotExist(non_existant_columns))
} else {
validate_condition(condition, schema)
let selection: ColumnSelectionForInterpreter =
columns.iter().filter_map(|column_name| schema.get_column_position(column_name)).collect();
let validated_condition = validate_condition(condition, schema)?;
Ok(OperationForInterpreter::Select(table_position, selection, validated_condition))
}
}
ColumnSelection::All => Ok(())
ColumnSelection::All => {
let validated_condition = validate_condition(condition, schema)?;
Ok(OperationForInterpreter::Select(table_position, schema.all_selection(), validated_condition))
}
}
}
// pub fn validate_update(table_name: &str, insertion_values: &InsertionValues, db_metadata: &Vec<(String, TableSchema)>) -> Result<(), ValidationError> {
// let schema = validate_table_exists(db_schema, table_name)?;
// let mut column_names = HashSet::new();
// // Find duplicate columns
// for (name, _) in insertion_values {
// if column_names.contains(name) {
// return Err(ValidationError::DuplicateColumn(name.clone()));
// } else {
// column_names.insert(name.clone());
// }
// }
// // Ensure columns exist in schema
// let column_value_type: Vec<_> = insertion_values.iter().map(|(column, value)| {
// (column, value, schema.column_name_position_mapping.iter().find(|(name, _) | {
// (*name).eq(column)
// }).map(|(_, t)| schema.types.get(*t as usize)))
// }).collect();
// if let Some((name, _, _)) = column_value_type.iter().find(|(_, _, t)| {
// t.is_none()
// }) {
// return Err(ValidationError::ColumnsDoNotExist(vec![(*name).clone())]);
// }
// // Check types
// if let Some((_, _, _)) = column_value_type.iter().find(|(_, value, t)| {
// if let Some(Some(column_type)) = t {
// !type_of(value).eq(column_type)
// } else {
// false
// }
// }) {
// // TODO: Add column name information
// return Err(ValidationError::TypeMismatch);
// }
// Ok(())
// }
pub fn validate_insert(table_name: &TableName, insertion_values: &InsertionValues, db_schema: &DbSchema) -> Result<(), ValidationError> {
let schema = validate_table_exists(db_schema, table_name)?;
pub fn validate_insert(table_name: TableName, insertion_values: InsertionValues, db_schema: &DbSchema) -> Result<OperationForInterpreter, ValidationError> {
let (table_position, schema) = validate_table_exists(db_schema, &table_name)?;
// Check for duplicate columns in insertion_values.
let columns_in_query_vec: Vec<&ColumnName> = insertion_values.iter().map(|(column_name, _)| column_name).collect();
@ -157,56 +112,64 @@ pub fn validate_insert(table_name: &TableName, insertion_values: &InsertionValue
let columns_in_schema: HashSet<&ColumnName> = HashSet::from_iter(schema.get_columns());
let non_existant_columns = Vec::from_iter(columns_in_query.difference(&columns_in_schema));
if non_existant_columns.len() > 0 {
return Err(ValidationError::ColumnsDoNotExist(non_existant_columns.iter().map(|str| str.to_string()).collect()));
return Err(ValidationError::ColumnsDoNotExist(non_existant_columns.iter().map(|column_name| column_name.to_string()).collect()));
}
let missing_required_columns = Vec::from_iter(columns_in_schema.difference(&columns_in_query));
if missing_required_columns.len() > 0 {
return Err(ValidationError::RequiredColumnsAreMissing(missing_required_columns.iter().map(|str| str.to_string()).collect()));
}
// Check types
// Check types and prepare for creation of InsertionValues for the interpreter
let mut values_map: HashMap<_, Value> = HashMap::new();
for (column_name, value) in insertion_values {
let expected_type = schema.get_type_at(column_name).ok_or(ValidationError::ColumnsDoNotExist(vec![column_name.to_string()]))?; // By the previous validation steps this is never gonna trigger an error.
let (column, expected_type) = schema.get_column0(&column_name).ok_or(ValidationError::ColumnsDoNotExist(vec![column_name.to_string()]))?; // By the previous validation steps this is never gonna trigger an error.
let value_type = value.to_type();
if value_type != expected_type {
return Err(ValidationError::TypeMismatch { column_name: column_name.to_string(), received_type: value_type, expected_type });
}
values_map.insert(column, value);
}
Ok(())
// These are values ordered by the column position
let values: InsertionValuesForInterpreter = values_map.into_values().collect();
Ok(OperationForInterpreter::Insert(table_position, values))
}
pub fn validate_delete(table_name: &TableName, condition: &Option<Condition>, db_schema: &DbSchema) -> Result<(), ValidationError> {
let schema = validate_table_exists(db_schema, table_name)?;
validate_condition(condition, schema)?;
Ok(())
pub fn validate_delete(table_name: TableName, condition: Option<Condition>, db_schema: &DbSchema) -> Result<OperationForInterpreter, ValidationError> {
let (table_position, schema) = validate_table_exists(db_schema, &table_name)?;
let validated_condition = validate_condition(condition, schema)?;
Ok(OperationForInterpreter::Delete(table_position, validated_condition))
}
fn validate_condition(condition: &Option<Condition>, schema: &TableSchema) -> Result<(), ValidationError> {
fn validate_condition(condition: Option<Condition>, schema: &TableSchema) -> Result<Option<ConditionForInterpreter>, ValidationError> {
match condition {
Some(condition) => {
match condition {
Condition::Eq(column_name, value) => {
let expected_type: DbType = schema.get_type_at(column_name).ok_or(ValidationError::ColumnsDoNotExist(vec![column_name.to_string()]))?;
let (column, expected_type) = schema.get_column0(&column_name).ok_or(ValidationError::ColumnsDoNotExist(vec![column_name.to_string()]))?;
let value_type: DbType = value.to_type();
if !expected_type.eq(&value_type) {
if expected_type.eq(&value_type) {
Ok(Some(ConditionForInterpreter::Eq(column, value)))
} else {
return Err(ValidationError::TypeMismatch { column_name: column_name.to_string(), received_type: value_type, expected_type });
}
}
}
}
None => {}
None => Ok(None)
}
Ok(())
}
fn validate_create_index(table_name: &TableName, column_name: &ColumnName, db_schema: &DbSchema) -> Result<(), ValidationError> {
let schema = validate_table_exists(db_schema, table_name)?;
if schema.does_column_exist(column_name) {
Ok(())
} else {
Err(ValidationError::ColumnsDoNotExist(vec![column_name.to_string()]))
}
fn validate_create_index(table_name: TableName, column_name: ColumnName, db_schema: &DbSchema) -> Result<OperationForInterpreter, ValidationError> {
// TODO: You should disallow indexing of Number columns.
let (table_position, schema) = validate_table_exists(db_schema, &table_name)?;
schema
.get_column_position(&column_name)
.map_or_else(
|| Err(ValidationError::ColumnsDoNotExist(vec![column_name.to_string()])),
|column| Ok(OperationForInterpreter::CreateIndex(table_position, column))
)
}
// ===Helpers===
@ -225,6 +188,6 @@ where A: Eq + std::hash::Hash
}
fn get_table_schema<'a>(db_schema: &DbSchema<'a>, table_name: &'a TableName) -> Option<&'a TableSchema> {
let (_, table_schema) = db_schema.iter().find(|(tname, _)| table_name.eq(tname))?;
let (_, _, table_schema) = db_schema.iter().find(|(tname, _, _)| table_name.eq(tname))?;
Some(table_schema)
}