Merge branch 'main' into dataframe

README.md

@@ -153,6 +153,133 @@ let zipped_matrix = a.zip(&b, |x, y| x + y);
 assert_eq!(zipped_matrix.data(), &[6.0, 8.0, 10.0, 12.0]);
 ```
 
+---
+
+## DataFrame Usage Example
+
+```rust
+use chrono::NaiveDate;
+use rustframe::dataframe::DataFrame;
+use rustframe::utils::{BDateFreq, BDatesList};
+use std::any::TypeId;
+use std::collections::HashMap;
+
+// Helper for NaiveDate
+fn d(y: i32, m: u32, d: u32) -> NaiveDate {
+    NaiveDate::from_ymd_opt(y, m, d).unwrap()
+}
+
+// Create a new DataFrame
+let mut df = DataFrame::new();
+
+// Add columns of different types
+df.add_column("col_int1", vec![1, 2, 3, 4, 5]);
+df.add_column("col_float1", vec![1.1, 2.2, 3.3, 4.4, 5.5]);
+df.add_column(
+    "col_string",
+    vec![
+        "apple".to_string(),
+        "banana".to_string(),
+        "cherry".to_string(),
+        "date".to_string(),
+        "elderberry".to_string(),
+    ],
+);
+df.add_column("col_bool", vec![true, false, true, false, true]);
+// df.add_column("col_date", vec![d(2023,1,1), d(2023,1,2), d(2023,1,3), d(2023,1,4), d(2023,1,5)]);
+df.add_column(
+    "col_date",
+    BDatesList::from_n_periods("2023-01-01".to_string(), BDateFreq::Daily, 5)
+        .unwrap()
+        .list()
+        .unwrap(),
+);
+
+println!("DataFrame after initial column additions:\n{}", df);
+
+// Demonstrate frame re-use when adding columns of existing types
+let initial_frames_count = df.num_internal_frames();
+println!(
+    "\nInitial number of internal frames: {}",
+    initial_frames_count
+);
+
+df.add_column("col_int2", vec![6, 7, 8, 9, 10]);
+df.add_column("col_float2", vec![6.6, 7.7, 8.8, 9.9, 10.0]);
+
+let frames_after_reuse = df.num_internal_frames();
+println!(
+    "Number of internal frames after adding more columns of existing types: {}",
+    frames_after_reuse
+);
+assert_eq!(initial_frames_count, frames_after_reuse); // Should be equal, demonstrating re-use
+
+println!(
+    "\nDataFrame after adding more columns of existing types:\n{}",
+    df
+);
+
+// Get number of rows and columns
+println!("Rows: {}", df.rows()); // Output: Rows: 5
+println!("Columns: {}", df.cols()); // Output: Columns: 5
+
+// Get column names
+println!("Column names: {:?}", df.get_column_names());
+// Output: Column names: ["col_int", "col_float", "col_string", "col_bool", "col_date"]
+
+// Get a specific column by name and type
+let int_col = df.get_column::<i32>("col_int1").unwrap();
+// Output: Integer column: [1, 2, 3, 4, 5]
+println!("Integer column (col_int1): {:?}", int_col);
+
+let int_col2 = df.get_column::<i32>("col_int2").unwrap();
+// Output: Integer column: [6, 7, 8, 9, 10]
+println!("Integer column (col_int2): {:?}", int_col2);
+
+let float_col = df.get_column::<f64>("col_float1").unwrap();
+// Output: Float column: [1.1, 2.2, 3.3, 4.4, 5.5]
+println!("Float column (col_float1): {:?}", float_col);
+
+// Attempt to get a column with incorrect type (returns None)
+let wrong_type_col = df.get_column::<bool>("col_int1");
+// Output: Wrong type column: None
+println!("Wrong type column: {:?}", wrong_type_col);
+
+// Get a row by index
+let row_0 = df.get_row(0).unwrap();
+println!("Row 0: {:?}", row_0);
+// Output: Row 0: {"col_int1": "1", "col_float1": "1.1", "col_string": "apple", "col_bool": "true", "col_date": "2023-01-01", "col_int2": "6", "col_float2": "6.6"}
+
+let row_2 = df.get_row(2).unwrap();
+println!("Row 2: {:?}", row_2);
+// Output: Row 2: {"col_int1": "3", "col_float1": "3.3", "col_string": "cherry", "col_bool": "true", "col_date": "2023-01-03", "col_int2": "8", "col_float2": "8.8"}
+
+// Attempt to get an out-of-bounds row (returns None)
+let row_out_of_bounds = df.get_row(10);
+// Output: Row out of bounds: None
+println!("Row out of bounds: {:?}", row_out_of_bounds);
+
+// Drop a column
+df.drop_column("col_bool");
+println!("\nDataFrame after dropping 'col_bool':\n{}", df);
+
+println!("Columns after drop: {}", df.cols());
+println!("Column names after drop: {:?}", df.get_column_names());
+
+// Drop another column, ensuring the underlying Frame is removed if empty
+df.drop_column("col_float1");
+println!("\nDataFrame after dropping 'col_float1':\n{}", df);
+
+println!("Columns after second drop: {}", df.cols());
+println!(
+    "Column names after second drop: {:?}",
+    df.get_column_names()
+);
+
+// Attempt to drop a non-existent column (will panic)
+// df.drop_column("non_existent_col"); // Uncomment to see panic
+```
+
 ## More examples
 
 See the [examples](./examples/) directory for some demonstrations of Rustframe's syntax and functionality.

src/csv/csv_core.rs

@@ -1,411 +0,0 @@
-use chrono::{NaiveDate, NaiveDateTime};
-use std::collections::HashMap;
-use std::fs::File;
-use std::io::{self, BufRead, BufReader};
-use std::path::Path;
-
-/// Represents the target type for a CSV column.
-#[derive(Debug, Clone)]
-pub enum DataType {
-    Int,
-    Float,
-    Bool,
-    UInt,
-    String,
-    Date,
-    DateTime,
-}
-
-/// Represents a value parsed from the CSV.
-#[derive(Debug, Clone, PartialEq)]
-pub enum Value {
-    Int(i64),
-    Float(f64),
-    Bool(bool),
-    UInt(u64),
-    String(String),
-    Date(NaiveDate),
-    DateTime(NaiveDateTime),
-}
-
-/// Convenience alias for a parsed CSV record.
-pub type Record = HashMap<String, Value>;
-
-/// A simple CSV reader that reads records line by line.
-pub struct CsvReader<R: BufRead> {
-    reader: R,
-    separators: Vec<char>,
-    headers: Vec<String>,
-    types: Option<HashMap<String, DataType>>,
-}
-
-/// Builder for [`CsvReader`] allowing chained configuration of headers, types, and separators.
-pub struct CsvReaderBuilder<R: BufRead> {
-    reader: R,
-    separators: Vec<char>,
-    headers: Vec<String>,
-    types: Option<HashMap<String, DataType>>,
-}
-
-impl<R: BufRead> CsvReader<R> {
-    /// Create a new CSV reader from a [`BufRead`] source.
-    /// The first line is expected to contain headers.
-    /// `separators` is a list of characters considered as field separators.
-    /// `types` optionally maps column names to target data types.
-    pub fn new(
-        mut reader: R,
-        separators: Vec<char>,
-        types: Option<HashMap<String, DataType>>,
-    ) -> io::Result<Self> {
-        let mut first_line = String::new();
-        reader.read_line(&mut first_line)?;
-        let headers = parse_line(&first_line, &separators);
-        Ok(Self {
-            reader,
-            separators,
-            headers,
-            types,
-        })
-    }
-
-    /// Create a reader with default settings (comma separator, automatic typing).
-    pub fn new_default(reader: R) -> io::Result<Self> {
-        Self::new(reader, vec![','], None)
-    }
-
-    /// Create a reader with default separators and explicit type mapping.
-    pub fn new_with_types(reader: R, types: HashMap<String, DataType>) -> io::Result<Self> {
-        Self::new(reader, vec![','], Some(types))
-    }
-
-    /// Start building a reader from a source that lacks headers.
-    pub fn new_with_headers(reader: R, headers: Vec<String>) -> CsvReaderBuilder<R> {
-        CsvReaderBuilder {
-            reader,
-            separators: vec![','],
-            headers,
-            types: None,
-        }
-    }
-
-    /// Return the headers of the CSV file.
-    pub fn headers(&self) -> &[String] {
-        &self.headers
-    }
-
-    /// Read the next record. Returns `Ok(None)` on EOF.
-    pub fn read_record(&mut self) -> io::Result<Option<Record>> {
-        let mut line = String::new();
-        if self.reader.read_line(&mut line)? == 0 {
-            return Ok(None);
-        }
-        let fields = parse_line(&line, &self.separators);
-        let mut record = HashMap::new();
-
-        for (i, header) in self.headers.iter().enumerate() {
-            let field = fields.get(i).cloned().unwrap_or_default();
-            let value = match &self.types {
-                Some(map) => {
-                    if let Some(dt) = map.get(header) {
-                        parse_with_type(&field, dt)
-                    } else {
-                        Value::String(field)
-                    }
-                }
-                None => parse_auto(&field),
-            };
-            record.insert(header.clone(), value);
-        }
-
-        Ok(Some(record))
-    }
-}
-
-impl<R: BufRead> Iterator for CsvReader<R> {
-    type Item = io::Result<Record>;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        match self.read_record() {
-            Ok(Some(rec)) => Some(Ok(rec)),
-            Ok(None) => None,
-            Err(e) => Some(Err(e)),
-        }
-    }
-}
-
-impl<R: BufRead> CsvReaderBuilder<R> {
-    /// Override field separators for the upcoming reader.
-    pub fn separators(mut self, separators: Vec<char>) -> Self {
-        self.separators = separators;
-        self
-    }
-
-    /// Finalize the builder with an explicit type mapping.
-    pub fn new_with_types(mut self, types: HashMap<String, DataType>) -> CsvReader<R> {
-        self.types = Some(types);
-        self.build()
-    }
-
-    /// Finalize the builder without specifying types.
-    pub fn build(self) -> CsvReader<R> {
-        CsvReader {
-            reader: self.reader,
-            separators: self.separators,
-            headers: self.headers,
-            types: self.types,
-        }
-    }
-}
-
-impl<R: BufRead> CsvReader<R> {
-    /// Read all remaining records into a vector.
-    pub fn read_all(&mut self) -> io::Result<Vec<Record>> {
-        let mut records = Vec::new();
-        while let Some(rec) = self.read_record()? {
-            records.push(rec);
-        }
-        Ok(records)
-    }
-}
-
-impl CsvReader<BufReader<File>> {
-    /// Create a [`CsvReader`] from a file path using comma separators and
-    /// automatic type detection.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use rustframe::csv::{CsvReader, Value};
-    /// # let path = std::env::temp_dir().join("from_path_auto.csv");
-    /// # std::fs::write(&path, "a,b\n1,true\n").unwrap();
-    /// let mut reader = CsvReader::from_path_auto(&path).unwrap();
-    /// let rec = reader.next().unwrap().unwrap();
-    /// assert_eq!(rec.get("a"), Some(&Value::Int(1)));
-    /// assert_eq!(rec.get("b"), Some(&Value::Bool(true)));
-    /// # std::fs::remove_file(path).unwrap();
-    /// ```
-    pub fn from_path_auto<P: AsRef<Path>>(path: P) -> io::Result<Self> {
-        let file = File::open(path)?;
-        let reader = BufReader::new(file);
-        CsvReader::new_default(reader)
-    }
-}
-
-/// Create an iterator over records from a file path using default settings.
-pub fn reader<P: AsRef<Path>>(path: P) -> io::Result<CsvReader<BufReader<File>>> {
-    reader_with(path, vec![','], None)
-}
-
-/// Create an iterator over records from a file path with custom separators and type mapping.
-pub fn reader_with<P: AsRef<Path>>(
-    path: P,
-    separators: Vec<char>,
-    types: Option<HashMap<String, DataType>>,
-) -> io::Result<CsvReader<BufReader<File>>> {
-    let file = File::open(path)?;
-    let reader = BufReader::new(file);
-    CsvReader::new(reader, separators, types)
-}
-
-/// Read an entire CSV file into memory using default settings.
-pub fn read_file<P: AsRef<Path>>(path: P) -> io::Result<Vec<Record>> {
-    read_file_with(path, vec![','], None)
-}
-
-/// Read an entire CSV file into memory with custom separators and type mapping.
-pub fn read_file_with<P: AsRef<Path>>(
-    path: P,
-    separators: Vec<char>,
-    types: Option<HashMap<String, DataType>>,
-) -> io::Result<Vec<Record>> {
-    let mut reader = reader_with(path, separators, types)?;
-    reader.read_all()
-}
-
-fn parse_with_type(s: &str, ty: &DataType) -> Value {
-    match ty {
-        DataType::Int => s
-            .parse::<i64>()
-            .map(Value::Int)
-            .unwrap_or_else(|_| Value::String(s.to_string())),
-        DataType::Float => s
-            .parse::<f64>()
-            .map(Value::Float)
-            .unwrap_or_else(|_| Value::String(s.to_string())),
-        DataType::Bool => s
-            .parse::<bool>()
-            .map(Value::Bool)
-            .unwrap_or_else(|_| Value::String(s.to_string())),
-        DataType::UInt => s
-            .parse::<u64>()
-            .map(Value::UInt)
-            .unwrap_or_else(|_| Value::String(s.to_string())),
-        DataType::String => Value::String(s.to_string()),
-        DataType::Date => s
-            .parse::<NaiveDate>()
-            .map(Value::Date)
-            .unwrap_or_else(|_| Value::String(s.to_string())),
-        DataType::DateTime => NaiveDateTime::parse_from_str(s, "%Y-%m-%d %H:%M:%S")
-            .map(Value::DateTime)
-            .unwrap_or_else(|_| Value::String(s.to_string())),
-    }
-}
-
-fn parse_auto(s: &str) -> Value {
-    if let Ok(i) = s.parse::<i64>() {
-        Value::Int(i)
-    } else if let Ok(f) = s.parse::<f64>() {
-        Value::Float(f)
-    } else if let Ok(b) = s.parse::<bool>() {
-        Value::Bool(b)
-    } else if let Ok(dt) = NaiveDateTime::parse_from_str(s, "%Y-%m-%d %H:%M:%S") {
-        Value::DateTime(dt)
-    } else if let Ok(d) = NaiveDate::parse_from_str(s, "%Y-%m-%d") {
-        Value::Date(d)
-    } else {
-        Value::String(s.to_string())
-    }
-}
-
-fn parse_line(line: &str, separators: &[char]) -> Vec<String> {
-    let mut fields = Vec::new();
-    let mut current = String::new();
-    let mut in_quotes: Option<char> = None;
-    let chars: Vec<char> = line.chars().collect();
-    let mut i = 0;
-
-    while i < chars.len() {
-        let c = chars[i];
-        if let Some(q) = in_quotes {
-            if c == q {
-                if i + 1 < chars.len() && chars[i + 1] == q {
-                    current.push(q);
-                    i += 1; // skip escaped quote
-                } else {
-                    in_quotes = None;
-                }
-            } else {
-                current.push(c);
-            }
-        } else if c == '"' || c == '\'' {
-            in_quotes = Some(c);
-        } else if separators.contains(&c) {
-            fields.push(current.clone());
-            current.clear();
-        } else if c == '\r' {
-            // Ignore carriage returns
-        } else if c == '\n' {
-            break;
-        } else {
-            current.push(c);
-        }
-        i += 1;
-    }
-
-    fields.push(current);
-    fields
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use chrono::{NaiveDate, NaiveDateTime};
-    use std::io::Cursor;
-
-    #[test]
-    fn test_parse_line() {
-        let line = "a,'b,c',\"d\"\"e\",f";
-        let fields = parse_line(line, &[',']);
-        assert_eq!(fields, vec!["a", "b,c", "d\"e", "f"]);
-    }
-
-    #[test]
-    fn test_reader_auto() {
-        let data = "a,b,c\n1,2.5,true\n4,5.0,false\n";
-        let cursor = Cursor::new(data);
-        let mut reader = CsvReader::new_default(cursor).unwrap();
-        let rec = reader.next().unwrap().unwrap();
-        assert_eq!(rec.get("a"), Some(&Value::Int(1)));
-        assert_eq!(rec.get("b"), Some(&Value::Float(2.5)));
-        assert_eq!(rec.get("c"), Some(&Value::Bool(true)));
-    }
-
-    #[test]
-    fn test_reader_with_types() {
-        let data = "a,b,c\n1,2,3\n";
-        let cursor = Cursor::new(data);
-        let mut types = HashMap::new();
-        types.insert("a".to_string(), DataType::Int);
-        types.insert("b".to_string(), DataType::Int);
-        types.insert("c".to_string(), DataType::String);
-        let mut reader = CsvReader::new_with_types(cursor, types).unwrap();
-        let rec = reader.next().unwrap().unwrap();
-        assert_eq!(rec.get("a"), Some(&Value::Int(1)));
-        assert_eq!(rec.get("b"), Some(&Value::Int(2)));
-        assert_eq!(rec.get("c"), Some(&Value::String("3".to_string())));
-    }
-
-    #[test]
-    fn test_chain_headers_and_types() {
-        let data = "1,2\n3,4\n";
-        let cursor = Cursor::new(data);
-        let headers = vec!["x".to_string(), "y".to_string()];
-        let mut types = HashMap::new();
-        types.insert("x".to_string(), DataType::Int);
-        types.insert("y".to_string(), DataType::UInt);
-        let mut reader = CsvReader::new_with_headers(cursor, headers).new_with_types(types);
-        let rec = reader.next().unwrap().unwrap();
-        assert_eq!(rec.get("x"), Some(&Value::Int(1)));
-        assert_eq!(rec.get("y"), Some(&Value::UInt(2)));
-    }
-
-    #[test]
-    fn test_date_types() {
-        let data = "d,dt\n2024-01-01,2024-01-01 12:00:00\n";
-        let cursor = Cursor::new(data);
-        let mut types = HashMap::new();
-        types.insert("d".to_string(), DataType::Date);
-        types.insert("dt".to_string(), DataType::DateTime);
-        let mut reader = CsvReader::new_with_types(cursor, types).unwrap();
-        let rec = reader.next().unwrap().unwrap();
-        let date = NaiveDate::from_ymd_opt(2024, 1, 1).unwrap();
-        let datetime: NaiveDateTime = NaiveDate::from_ymd_opt(2024, 1, 1)
-            .unwrap()
-            .and_hms_opt(12, 0, 0)
-            .unwrap();
-        assert_eq!(rec.get("d"), Some(&Value::Date(date)));
-        assert_eq!(rec.get("dt"), Some(&Value::DateTime(datetime)));
-    }
-
-    #[test]
-    fn test_read_file_all() {
-        let path = std::env::temp_dir().join("csv_full_test.csv");
-        std::fs::write(&path, "a,b\n1,2\n3,4\n").unwrap();
-        let records = read_file(&path).unwrap();
-        assert_eq!(records.len(), 2);
-        assert_eq!(records[1].get("b"), Some(&Value::Int(4)));
-        std::fs::remove_file(path).unwrap();
-    }
-
-    #[test]
-    fn test_reader_from_path() {
-        let path = std::env::temp_dir().join("csv_iter_test.csv");
-        std::fs::write(&path, "a,b\n5,6\n").unwrap();
-        let mut iter = reader(&path).unwrap();
-        let rec = iter.next().unwrap().unwrap();
-        assert_eq!(rec.get("a"), Some(&Value::Int(5)));
-        assert_eq!(rec.get("b"), Some(&Value::Int(6)));
-        std::fs::remove_file(path).unwrap();
-    }
-
-    #[test]
-    fn test_from_path_auto_method() {
-        let path = std::env::temp_dir().join("csv_method_auto.csv");
-        std::fs::write(&path, "a,b\n7,true\n").unwrap();
-        let mut reader = CsvReader::from_path_auto(&path).unwrap();
-        let rec = reader.next().unwrap().unwrap();
-        assert_eq!(rec.get("a"), Some(&Value::Int(7)));
-        assert_eq!(rec.get("b"), Some(&Value::Bool(true)));
-        std::fs::remove_file(path).unwrap();
-    }
-}

src/csv/mod.rs

@@ -1,69 +0,0 @@
-//! CSV handling utilities.
-//!
-//! The [`csv`] module offers a flexible [`CsvReader`] with automatic type
-//! detection and optional builders for custom headers and types.
-//!
-//! # Examples
-//!
-//! Read from a file with auto type detection:
-//!
-//! ```
-//! use rustframe::csv::CsvReader;
-//! # let path = std::env::temp_dir().join("docs_auto.csv");
-//! # std::fs::write(&path, "a,b\n1,true\n").unwrap();
-//! let mut reader = CsvReader::from_path_auto(&path).unwrap();
-//! for rec in reader {
-//!     let rec = rec.unwrap();
-//!     println!("{:?}", rec);
-//! }
-//! # std::fs::remove_file(path).unwrap();
-//! ```
-//!
-//! Specify column types explicitly:
-//!
-//! ```
-//! use rustframe::csv::{CsvReader, DataType, Value};
-//! use std::collections::HashMap;
-//! use std::io::Cursor;
-//! let data = "a,b\n1,2\n";
-//! let mut types = HashMap::new();
-//! types.insert("a".into(), DataType::Int);
-//! types.insert("b".into(), DataType::Float);
-//! let mut reader = CsvReader::new_with_types(Cursor::new(data), types).unwrap();
-//! let rec = reader.next().unwrap().unwrap();
-//! assert_eq!(rec.get("b"), Some(&Value::Float(2.0)));
-//! ```
-//!
-//! Building from custom headers and types:
-//!
-//! ```
-//! use rustframe::csv::{CsvReader, DataType, Value};
-//! use std::collections::HashMap;
-//! use std::io::Cursor;
-//! let data = "1,2\n";
-//! let headers = vec!["x".to_string(), "y".to_string()];
-//! let mut types = HashMap::new();
-//! types.insert("x".into(), DataType::Int);
-//! types.insert("y".into(), DataType::UInt);
-//! let mut reader = CsvReader::new_with_headers(Cursor::new(data), headers).new_with_types(types);
-//! let rec = reader.next().unwrap().unwrap();
-//! assert_eq!(rec.get("y"), Some(&Value::UInt(2)));
-//! ```
-//!
-//! Reading an entire file into memory:
-//!
-//! ```
-//! use rustframe::csv::read_file;
-//! # let path = std::env::temp_dir().join("docs_full.csv");
-//! # std::fs::write(&path, "a,b\n1,2\n3,4\n").unwrap();
-//! let records = read_file(&path).unwrap();
-//! assert_eq!(records.len(), 2);
-//! # std::fs::remove_file(path).unwrap();
-//! ```
-
-pub mod csv_core;
-
-pub use csv_core::{
-    CsvReader, CsvReaderBuilder, DataType, Record, Value, reader, reader_with,
-    read_file, read_file_with,
-};

src/dataframe/df.rs

@@ -0,0 +1,659 @@
+use crate::frame::{Frame, RowIndex};
+use std::any::{Any, TypeId};
+use std::collections::HashMap;
+use std::fmt; // Import TypeId
+
+const DEFAULT_DISPLAY_ROWS: usize = 5;
+const DEFAULT_DISPLAY_COLS: usize = 10;
+
+// Trait to enable type-agnostic operations on Frame objects within DataFrame
+pub trait SubFrame: Send + Sync + fmt::Debug + Any {
+    fn rows(&self) -> usize;
+    fn get_value_as_string(&self, physical_row_idx: usize, col_name: &str) -> String;
+    fn clone_box(&self) -> Box<dyn SubFrame>;
+    fn delete_column_from_frame(&mut self, col_name: &str);
+    fn get_frame_cols(&self) -> usize; // Add a method to get the number of columns in the underlying frame
+
+    // Methods for downcasting to concrete types
+    fn as_any(&self) -> &dyn Any;
+    fn as_any_mut(&mut self) -> &mut dyn Any;
+}
+
+// Implement SubFrame for any Frame<T> that meets the requirements
+impl<T> SubFrame for Frame<T>
+where
+    T: Clone + PartialEq + fmt::Display + fmt::Debug + 'static + Send + Sync + Any,
+{
+    fn rows(&self) -> usize {
+        self.rows()
+    }
+
+    fn get_value_as_string(&self, physical_row_idx: usize, col_name: &str) -> String {
+        self.get_row(physical_row_idx).get(col_name).to_string()
+    }
+
+    fn clone_box(&self) -> Box<dyn SubFrame> {
+        Box::new(self.clone())
+    }
+
+    fn delete_column_from_frame(&mut self, col_name: &str) {
+        self.delete_column(col_name);
+    }
+
+    fn get_frame_cols(&self) -> usize {
+        self.cols()
+    }
+
+    fn as_any(&self) -> &dyn Any {
+        self
+    }
+
+    fn as_any_mut(&mut self) -> &mut dyn Any {
+        self
+    }
+}
+
+pub struct DataFrame {
+    frames_by_type: HashMap<TypeId, Box<dyn SubFrame>>, // Maps TypeId to the Frame holding columns of that type
+    column_to_type: HashMap<String, TypeId>,            // Maps column name to its TypeId
+    column_names: Vec<String>,
+    index: RowIndex,
+}
+
+impl DataFrame {
+    pub fn new() -> Self {
+        DataFrame {
+            frames_by_type: HashMap::new(),
+            column_to_type: HashMap::new(),
+            column_names: Vec::new(),
+            index: RowIndex::Range(0..0), // Initialize with an empty range index
+        }
+    }
+
+    /// Returns the number of rows in the DataFrame.
+    pub fn rows(&self) -> usize {
+        self.index.len()
+    }
+
+    /// Returns the number of columns in the DataFrame.
+    pub fn cols(&self) -> usize {
+        self.column_names.len()
+    }
+
+    /// Returns a reference to the vector of column names.
+    pub fn get_column_names(&self) -> &Vec<String> {
+        &self.column_names
+    }
+
+    /// Returns the number of internal Frame objects (one per unique data type).
+    pub fn num_internal_frames(&self) -> usize {
+        self.frames_by_type.len()
+    }
+
+    /// Returns a reference to a column of a specific type, if it exists.
+    pub fn get_column<T>(&self, col_name: &str) -> Option<&[T]>
+    where
+        T: Clone + PartialEq + fmt::Display + fmt::Debug + 'static + Send + Sync + Any,
+    {
+        let expected_type_id = TypeId::of::<T>();
+        if let Some(actual_type_id) = self.column_to_type.get(col_name) {
+            if *actual_type_id == expected_type_id {
+                if let Some(sub_frame_box) = self.frames_by_type.get(actual_type_id) {
+                    if let Some(frame) = sub_frame_box.as_any().downcast_ref::<Frame<T>>() {
+                        return Some(frame.column(col_name));
+                    }
+                }
+            }
+        }
+        None
+    }
+
+    /// Returns a HashMap representing a row, mapping column names to their string values.
+    pub fn get_row(&self, row_idx: usize) -> Option<HashMap<String, String>> {
+        if row_idx >= self.rows() {
+            return None;
+        }
+
+        let mut row_data = HashMap::new();
+        for col_name in &self.column_names {
+            if let Some(type_id) = self.column_to_type.get(col_name) {
+                if let Some(sub_frame_box) = self.frames_by_type.get(type_id) {
+                    let value = sub_frame_box.get_value_as_string(row_idx, col_name);
+                    row_data.insert(col_name.clone(), value);
+                }
+            }
+        }
+        Some(row_data)
+    }
+
+    pub fn add_column<T>(&mut self, col_name: &str, data: Vec<T>)
+    where
+        T: Clone + PartialEq + fmt::Display + fmt::Debug + 'static + Send + Sync + Any,
+    {
+        let type_id = TypeId::of::<T>();
+        let col_name_string = col_name.to_string();
+
+        // Check for duplicate column name across the entire DataFrame
+        if self.column_to_type.contains_key(&col_name_string) {
+            panic!(
+                "DataFrame::add_column: duplicate column name: '{}'",
+                col_name_string
+            );
+        }
+
+        // If this is the first column being added, set the DataFrame's index
+        if self.column_names.is_empty() {
+            self.index = RowIndex::Range(0..data.len());
+        } else {
+            // Ensure new column has the same number of rows as existing columns
+            if data.len() != self.index.len() {
+                panic!(
+                    "DataFrame::add_column: new column '{}' has {} rows, but existing columns have {} rows",
+                    col_name_string,
+                    data.len(),
+                    self.index.len()
+                );
+            }
+        }
+
+        // Check if a Frame of this type already exists
+        if let Some(sub_frame_box) = self.frames_by_type.get_mut(&type_id) {
+            // Downcast to the concrete Frame<T> and add the column
+            if let Some(frame) = sub_frame_box.as_any_mut().downcast_mut::<Frame<T>>() {
+                frame.add_column(col_name_string.clone(), data);
+            } else {
+                // This should ideally not happen if TypeId matches, but good for safety
+                panic!(
+                    "Type mismatch when downcasting existing SubFrame for TypeId {:?}",
+                    type_id
+                );
+            }
+        } else {
+            // No Frame of this type exists, create a new one
+            // The Frame::new constructor expects a Matrix and column names.
+            // We create a Matrix from a single column vector.
+            let new_frame = Frame::new(
+                crate::matrix::Matrix::from_cols(vec![data]),
+                vec![col_name_string.clone()],
+                Some(self.index.clone()), // Pass the DataFrame's index to the new Frame
+            );
+            self.frames_by_type.insert(type_id, Box::new(new_frame));
+        }
+
+        // Update column mappings and names
+        self.column_to_type.insert(col_name_string.clone(), type_id);
+        self.column_names.push(col_name_string);
+    }
+
+    /// Drops a column from the DataFrame.
+    /// Panics if the column does not exist.
+    pub fn drop_column(&mut self, col_name: &str) {
+        let col_name_string = col_name.to_string();
+
+        // 1. Get the TypeId associated with the column
+        let type_id = self
+            .column_to_type
+            .remove(&col_name_string)
+            .unwrap_or_else(|| {
+                panic!(
+                    "DataFrame::drop_column: column '{}' not found",
+                    col_name_string
+                );
+            });
+
+        // 2. Remove the column name from the ordered list
+        self.column_names.retain(|name| name != &col_name_string);
+
+        // 3. Find the Frame object and delete the column from it
+        if let Some(sub_frame_box) = self.frames_by_type.get_mut(&type_id) {
+            sub_frame_box.delete_column_from_frame(&col_name_string);
+
+            // 4. If the Frame object for this type becomes empty, remove it from frames_by_type
+            if sub_frame_box.get_frame_cols() == 0 {
+                self.frames_by_type.remove(&type_id);
+            }
+        } else {
+            // This should not happen if column_to_type was consistent
+            panic!(
+                "DataFrame::drop_column: internal error, no frame found for type_id {:?}",
+                type_id
+            );
+        }
+    }
+}
+
+impl fmt::Display for DataFrame {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        // Display column headers
+        for col_name in self.column_names.iter().take(DEFAULT_DISPLAY_COLS) {
+            write!(f, "{:<15}", col_name)?;
+        }
+        if self.column_names.len() > DEFAULT_DISPLAY_COLS {
+            write!(f, "...")?;
+        }
+        writeln!(f)?;
+
+        // Display data rows
+        let mut displayed_rows = 0;
+        for i in 0..self.index.len() {
+            if displayed_rows >= DEFAULT_DISPLAY_ROWS {
+                writeln!(f, "...")?;
+                break;
+            }
+            for col_name in self.column_names.iter().take(DEFAULT_DISPLAY_COLS) {
+                if let Some(type_id) = self.column_to_type.get(col_name) {
+                    if let Some(sub_frame_box) = self.frames_by_type.get(type_id) {
+                        write!(f, "{:<15}", sub_frame_box.get_value_as_string(i, col_name))?;
+                    } else {
+                        // This case indicates an inconsistency: column_to_type has an entry,
+                        // but frames_by_type doesn't have the corresponding Frame.
+                        write!(f, "{:<15}", "[ERROR]")?;
+                    }
+                } else {
+                    // This case indicates an inconsistency: column_names has an entry,
+                    // but column_to_type doesn't have the corresponding column.
+                    write!(f, "{:<15}", "[ERROR]")?;
+                }
+            }
+            if self.column_names.len() > DEFAULT_DISPLAY_COLS {
+                write!(f, "...")?;
+            }
+            writeln!(f)?;
+            displayed_rows += 1;
+        }
+        Ok(())
+    }
+}
+
+impl fmt::Debug for DataFrame {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        f.debug_struct("DataFrame")
+            .field("column_names", &self.column_names)
+            .field("index", &self.index)
+            .field("column_to_type", &self.column_to_type)
+            .field("frames_by_type", &self.frames_by_type)
+            .finish()
+    }
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::frame::Frame;
+    use crate::matrix::Matrix;
+
+    #[test]
+    fn test_dataframe_new() {
+        let df = DataFrame::new();
+        assert_eq!(df.rows(), 0);
+        assert_eq!(df.cols(), 0);
+        assert!(df.get_column_names().is_empty());
+        assert!(df.frames_by_type.is_empty());
+        assert!(df.column_to_type.is_empty());
+    }
+
+    #[test]
+    fn test_dataframe_add_column_initial() {
+        let mut df = DataFrame::new();
+        let data = vec![1, 2, 3];
+        df.add_column("col_int", data.clone());
+
+        assert_eq!(df.rows(), 3);
+        assert_eq!(df.cols(), 1);
+        assert_eq!(df.get_column_names(), &vec!["col_int".to_string()]);
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert_eq!(df.column_to_type.get("col_int"), Some(&TypeId::of::<i32>()));
+
+        // Verify the underlying frame
+        let sub_frame_box = df.frames_by_type.get(&TypeId::of::<i32>()).unwrap();
+        let frame = sub_frame_box.as_any().downcast_ref::<Frame<i32>>().unwrap();
+        assert_eq!(frame.rows(), 3);
+        assert_eq!(frame.cols(), 1);
+        assert_eq!(frame.columns(), &vec!["col_int".to_string()]);
+    }
+
+    #[test]
+    fn test_dataframe_add_column_same_type() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int1", vec![1, 2, 3]);
+        df.add_column("col_int2", vec![4, 5, 6]);
+
+        assert_eq!(df.rows(), 3);
+        assert_eq!(df.cols(), 2);
+        assert_eq!(
+            df.get_column_names(),
+            &vec!["col_int1".to_string(), "col_int2".to_string()]
+        );
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert_eq!(
+            df.column_to_type.get("col_int1"),
+            Some(&TypeId::of::<i32>())
+        );
+        assert_eq!(
+            df.column_to_type.get("col_int2"),
+            Some(&TypeId::of::<i32>())
+        );
+
+        // Verify the underlying frame
+        let sub_frame_box = df.frames_by_type.get(&TypeId::of::<i32>()).unwrap();
+        let frame = sub_frame_box.as_any().downcast_ref::<Frame<i32>>().unwrap();
+        assert_eq!(frame.rows(), 3);
+        assert_eq!(frame.cols(), 2);
+        assert_eq!(
+            frame.columns(),
+            &vec!["col_int1".to_string(), "col_int2".to_string()]
+        );
+    }
+
+    #[test]
+    fn test_dataframe_add_column_different_type() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3]);
+        df.add_column(
+            "col_string",
+            vec!["a".to_string(), "b".to_string(), "c".to_string()],
+        );
+
+        assert_eq!(df.rows(), 3);
+        assert_eq!(df.cols(), 3);
+        assert_eq!(
+            df.get_column_names(),
+            &vec![
+                "col_int".to_string(),
+                "col_float".to_string(),
+                "col_string".to_string()
+            ]
+        );
+
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<String>()));
+
+        assert_eq!(df.column_to_type.get("col_int"), Some(&TypeId::of::<i32>()));
+        assert_eq!(
+            df.column_to_type.get("col_float"),
+            Some(&TypeId::of::<f64>())
+        );
+        assert_eq!(
+            df.column_to_type.get("col_string"),
+            Some(&TypeId::of::<String>())
+        );
+
+        // Verify underlying frames
+        let int_frame = df
+            .frames_by_type
+            .get(&TypeId::of::<i32>())
+            .unwrap()
+            .as_any()
+            .downcast_ref::<Frame<i32>>()
+            .unwrap();
+        assert_eq!(int_frame.columns(), &vec!["col_int".to_string()]);
+
+        let float_frame = df
+            .frames_by_type
+            .get(&TypeId::of::<f64>())
+            .unwrap()
+            .as_any()
+            .downcast_ref::<Frame<f64>>()
+            .unwrap();
+        assert_eq!(float_frame.columns(), &vec!["col_float".to_string()]);
+
+        let string_frame = df
+            .frames_by_type
+            .get(&TypeId::of::<String>())
+            .unwrap()
+            .as_any()
+            .downcast_ref::<Frame<String>>()
+            .unwrap();
+        assert_eq!(string_frame.columns(), &vec!["col_string".to_string()]);
+    }
+
+    #[test]
+    fn test_dataframe_get_column() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3]);
+        df.add_column(
+            "col_string",
+            vec!["a".to_string(), "b".to_string(), "c".to_string()],
+        );
+
+        // Test getting existing columns with correct type
+        assert_eq!(
+            df.get_column::<i32>("col_int").unwrap(),
+            vec![1, 2, 3].as_slice()
+        );
+        assert_eq!(
+            df.get_column::<f64>("col_float").unwrap(),
+            vec![1.1, 2.2, 3.3].as_slice()
+        );
+        assert_eq!(
+            df.get_column::<String>("col_string").unwrap(),
+            vec!["a".to_string(), "b".to_string(), "c".to_string()].as_slice()
+        );
+
+        // Test getting non-existent column
+        assert_eq!(df.get_column::<i32>("non_existent"), None);
+
+        // Test getting existing column with incorrect type
+        assert_eq!(df.get_column::<f64>("col_int"), None);
+        assert_eq!(df.get_column::<i32>("col_float"), None);
+    }
+
+    #[test]
+    fn test_dataframe_get_row() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3]);
+        df.add_column(
+            "col_string",
+            vec!["a".to_string(), "b".to_string(), "c".to_string()],
+        );
+
+        // Test getting an existing row
+        let row0 = df.get_row(0).unwrap();
+        assert_eq!(row0.get("col_int"), Some(&"1".to_string()));
+        assert_eq!(row0.get("col_float"), Some(&"1.1".to_string()));
+        assert_eq!(row0.get("col_string"), Some(&"a".to_string()));
+
+        let row1 = df.get_row(1).unwrap();
+        assert_eq!(row1.get("col_int"), Some(&"2".to_string()));
+        assert_eq!(row1.get("col_float"), Some(&"2.2".to_string()));
+        assert_eq!(row1.get("col_string"), Some(&"b".to_string()));
+
+        // Test getting an out-of-bounds row
+        assert_eq!(df.get_row(3), None);
+    }
+
+    #[test]
+    #[should_panic(expected = "DataFrame::add_column: duplicate column name: 'col_int'")]
+    fn test_dataframe_add_column_duplicate_name() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.add_column("col_int", vec![4, 5, 6]);
+    }
+
+    #[test]
+    #[should_panic(
+        expected = "DataFrame::add_column: new column 'col_int2' has 2 rows, but existing columns have 3 rows"
+    )]
+    fn test_dataframe_add_column_mismatched_rows() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int1", vec![1, 2, 3]);
+        df.add_column("col_int2", vec![4, 5]);
+    }
+
+    #[test]
+    fn test_dataframe_display() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3, 4, 5, 6]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3, 4.4, 5.5, 6.6]);
+        df.add_column(
+            "col_string",
+            vec![
+                "a".to_string(),
+                "b".to_string(),
+                "c".to_string(),
+                "d".to_string(),
+                "e".to_string(),
+                "f".to_string(),
+            ],
+        );
+
+        let expected_output = "\
+col_int        col_float      col_string     
+1              1.1            a              
+2              2.2            b              
+3              3.3            c              
+4              4.4            d              
+5              5.5            e              
+...
+";
+        assert_eq!(format!("{}", df), expected_output);
+    }
+
+    #[test]
+    fn test_dataframe_debug() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3]);
+
+        let debug_output = format!("{:?}", df);
+        assert!(debug_output.contains("DataFrame {"));
+        assert!(debug_output.contains("column_names: [\"col_int\", \"col_float\"]"));
+        assert!(debug_output.contains("index: Range(0..3)"));
+        assert!(debug_output.contains("column_to_type: {"));
+        assert!(debug_output.contains("frames_by_type: {"));
+    }
+
+    #[test]
+    fn test_dataframe_drop_column_single_type() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int1", vec![1, 2, 3]);
+        df.add_column("col_int2", vec![4, 5, 6]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3]);
+
+        assert_eq!(df.cols(), 3);
+        assert_eq!(
+            df.get_column_names(),
+            &vec![
+                "col_int1".to_string(),
+                "col_int2".to_string(),
+                "col_float".to_string()
+            ]
+        );
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+
+        df.drop_column("col_int1");
+
+        assert_eq!(df.cols(), 2);
+        assert_eq!(
+            df.get_column_names(),
+            &vec!["col_int2".to_string(), "col_float".to_string()]
+        );
+        assert!(df.column_to_type.get("col_int1").is_none());
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>())); // Frame<i32> should still exist
+        let int_frame = df
+            .frames_by_type
+            .get(&TypeId::of::<i32>())
+            .unwrap()
+            .as_any()
+            .downcast_ref::<Frame<i32>>()
+            .unwrap();
+        assert_eq!(int_frame.columns(), &vec!["col_int2".to_string()]);
+
+        df.drop_column("col_int2");
+
+        assert_eq!(df.cols(), 1);
+        assert_eq!(df.get_column_names(), &vec!["col_float".to_string()]);
+        assert!(df.column_to_type.get("col_int2").is_none());
+        assert!(!df.frames_by_type.contains_key(&TypeId::of::<i32>())); // Frame<i32> should be removed
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+    }
+
+    #[test]
+    fn test_dataframe_drop_column_mixed_types() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.add_column("col_float", vec![1.1, 2.2, 3.3]);
+        df.add_column(
+            "col_string",
+            vec!["a".to_string(), "b".to_string(), "c".to_string()],
+        );
+
+        assert_eq!(df.cols(), 3);
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<String>()));
+
+        df.drop_column("col_float");
+
+        assert_eq!(df.cols(), 2);
+        assert_eq!(
+            df.get_column_names(),
+            &vec!["col_int".to_string(), "col_string".to_string()]
+        );
+        assert!(df.column_to_type.get("col_float").is_none());
+        assert!(!df.frames_by_type.contains_key(&TypeId::of::<f64>())); // Frame<f64> should be removed
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<String>()));
+
+        df.drop_column("col_int");
+        df.drop_column("col_string");
+
+        assert_eq!(df.cols(), 0);
+        assert!(df.get_column_names().is_empty());
+        assert!(df.frames_by_type.is_empty());
+        assert!(df.column_to_type.is_empty());
+    }
+
+    #[test]
+    #[should_panic(expected = "DataFrame::drop_column: column 'non_existent' not found")]
+    fn test_dataframe_drop_column_non_existent() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int", vec![1, 2, 3]);
+        df.drop_column("non_existent");
+    }
+
+    #[test]
+    fn test_dataframe_add_column_reuses_existing_frame() {
+        let mut df = DataFrame::new();
+        df.add_column("col_int1", vec![1, 2, 3]);
+        df.add_column("col_float1", vec![1.1, 2.2, 3.3]);
+
+        // Initially, there should be two frames (one for i32, one for f64)
+        assert_eq!(df.frames_by_type.len(), 2);
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+
+        // Add another integer column
+        df.add_column("col_int2", vec![4, 5, 6]);
+
+        // The number of frames should still be 2, as the existing i32 frame should be reused
+        assert_eq!(df.frames_by_type.len(), 2);
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+
+        // Verify the i32 frame now contains both integer columns
+        let int_frame = df.frames_by_type.get(&TypeId::of::<i32>()).unwrap().as_any().downcast_ref::<Frame<i32>>().unwrap();
+        assert_eq!(int_frame.columns(), &vec!["col_int1".to_string(), "col_int2".to_string()]);
+        assert_eq!(int_frame.cols(), 2);
+
+        // Add another float column
+        df.add_column("col_float2", vec![4.4, 5.5, 6.6]);
+
+        // The number of frames should still be 2, as the existing f64 frame should be reused
+        assert_eq!(df.frames_by_type.len(), 2);
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<i32>()));
+        assert!(df.frames_by_type.contains_key(&TypeId::of::<f64>()));
+
+        // Verify the f64 frame now contains both float columns
+        let float_frame = df.frames_by_type.get(&TypeId::of::<f64>()).unwrap().as_any().downcast_ref::<Frame<f64>>().unwrap();
+        assert_eq!(float_frame.columns(), &vec!["col_float1".to_string(), "col_float2".to_string()]);
+        assert_eq!(float_frame.cols(), 2);
+    }
+}

src/dataframe/mod.rs

@@ -0,0 +1,4 @@
+//! This module provides the DataFrame structure for handling tabular data with mixed types.
+pub mod df;
+
+pub use df::{DataFrame, SubFrame};

src/frame/base.rs

@@ -332,7 +332,7 @@ impl<T: Clone + PartialEq> Frame<T> {
         )
     }
 
-    /// Returns an immutable slice of the specified column's data.
+    /// Returns an immutable slice of the specified column's data by name.
     /// Panics if the column name is not found.
     pub fn column(&self, name: &str) -> &[T] {
         let idx = self
@@ -341,7 +341,13 @@ impl<T: Clone + PartialEq> Frame<T> {
         self.matrix.column(idx)
     }
 
-    /// Returns a mutable slice of the specified column's data.
+    /// Returns an immutable slice of the specified column's data by its physical index.
+    /// Panics if the index is out of bounds.
+    pub fn column_by_physical_idx(&self, idx: usize) -> &[T] {
+        self.matrix.column(idx)
+    }
+
+    /// Returns a mutable slice of the specified column's data by name.
     /// Panics if the column name is not found.
     pub fn column_mut(&mut self, name: &str) -> &mut [T] {
         let idx = self
@@ -350,6 +356,12 @@ impl<T: Clone + PartialEq> Frame<T> {
         self.matrix.column_mut(idx)
     }
 
+    /// Returns a mutable slice of the specified column's data by its physical index.
+    /// Panics if the index is out of bounds.
+    pub fn column_mut_by_physical_idx(&mut self, idx: usize) -> &mut [T] {
+        self.matrix.column_mut(idx)
+    }
+
     // Row access methods
 
     /// Returns an immutable view of the row for the given integer key.

src/lib.rs

@@ -1,5 +1,8 @@
 #![doc = include_str!("../README.md")]
 
+/// Documentation for the [`crate::dataframe`] module.
+pub mod dataframe;
+
 /// Documentation for the [`crate::matrix`] module.
 pub mod matrix;
 
@@ -14,6 +17,3 @@ pub mod compute;
 
 /// Documentation for the [`crate::random`] module.
 pub mod random;
-
-/// Documentation for the [`crate::csv`] module.
-pub mod csv;