DataFrame

DataFrame documentation with code samples

DataFrame is a high-performance C++ library for in-memory data exploration, transformation, and statistical analysis — designed for data scientists, quant traders, and C++ developers who need efficient tabular data processing without Python overhead. This library is designed to provide similar functionalities to data manipulation and analysis tools found in other languages, such as Python’s Pandas or R’s data.frame. It aims to offer a robust and efficient way to handle tabular data in C++. The depth and breadth of functionalities offered by C++ DataFrame alone are greater than functionalities offered by packages such as Pandas, data.frame, and Polars combined.

You can slice the data in many ways. You can join, merge, group-by, cross tabulate, pivot the data. You can run various statistical, summarization, financial, and ML algorithms on the data. You can add your custom algorithms easily. You can multi-column sort, custom pick and delete the data. And much more …

DataFrame also includes a large collection of analytical algorithms in form of visitors. These are from basic stats such as Mean, Stdev, Moving Averages, … to more involved analysis such as PCA, Polynomial Fit, FFT, Eigens … including a good collection of trading indicators. You can also easily add your own algorithms. Many of these algorithms work seamlessly with both scalar and multidimensional datasets.
DataFrame also employs extensive multithreading in almost all its API’s, for large datasets. That makes DataFrame especially suitable for analyzing large datasets.

For basic operations to start you off, see Hello World and/or Cheat Sheet. For a complete list of features with code samples, see documentation.

---

I have followed a few principles in this library:

1. Support any type either built-in or user defined without needing new code
2. Never chase pointers ala linked lists, std::any, pointer to base, …
3. Have all column data in contiguous memory space
4. Never use more space than you need ala unions, std::variant, …
5. Avoid copying data as much as possible
6. Use multi-threading but only when it makes sense
7. Do not attempt to protect the user against garbage in, garbage out
8. Keep DataFrame library self-contained, meaning DataFrame must only depend on C++ language and its standard library

---

Performance

You have probably heard of Polars DataFrame. It is implemented in Rust and ported with zero-overhead to Python (as long as you don’t have a loop). I have been asked by many people to write a comparison for DataFrame vs. Polars. So, I finally found some time to learn a bit about Polars and write a very simple benchmark.
I wrote the following identical programs for both Polars and C++ DataFrame (and Pandas). I used Polars version: 0.19.14 (Pandas version: 1.5.3, Numpy version: 1.24.2). And I used C++23 GCC-14 compiler with -O3 option. I ran both on my, somewhat outdated, MacBook Pro (Intel chip, 96GB RAM).
In both cases, I created a dataframe with 3 random columns. The C++ DataFrame also required an additional index column of the same size. Polars doesn’t believe in index columns (that has its own pros and cons. I am not going through it here). Each program has three identical parts. First it generates and populates 3 columns with 300m random numbers each (in case of C++ DataFrame, it must also generate a sequential index column of the same size). That is the part I am not interested in. In the second part, it calculates the mean of the first column, the variance of the second column, and the Pearson correlation of the second and third columns. In the third part, it does a select (or filter as Polars calls it) on one of the columns.

Results:
The maximum dataset I could load into Polars was 300m rows per column. Any bigger dataset blew up the memory and caused OS to kill it. I ran C++ DataFrame with 2b rows per column and I am sure it would have run with bigger datasets too. So, I was forced to run both with 300m rows to compare. I ran each test 4 times and took the best time. Polars numbers varied a lot from one run to another, especially calculation and selection times. C++ DataFrame numbers were significantly more consistent.

	C++ DataFrame	Polars	Pandas
Data Generation Time	26.9459 secs	28.4686 secs	36.6799 secs
Calculation Time	1.2602 secs	4.8766 secs	40.3264 secs
Selection Time	0.4215 secs	3.8766 secs	8.3264 secs
Overall Time	28.9486 secs	36.8763 secs	85.8451 secs

---

Please consider sponsoring DataFrame, especially if you are using it in production capacity. It is the strongest form of appreciation

---