Datasets:

WizzF
/

Heap-Forge

@@ -461,7 +461,7 @@ We develop **The Heap**, a new contamination-free multilingual code dataset comp
 We collect up to **50,000** public repositories using the [GitHub API](https://docs.github.com/en/rest/search/search?apiVersion=2022-11-28), focusing on *license type*, *star count*, and *creation date*. Repositories with non-permissive licenses are prioritized to reduce contamination, as public code datasets we deduplicate against primarily focus on permissive or no-license repositories. We select repositories created before **April 2024** in decreasing order of their star counts. To handle GitHub rate limits, we use timeouts and pagination during the scraping process.
-### Copyleft licenses included in the dataset
 | **License**    | **Family** |
 | :--------: | :-------: |
 | CECILL-1.0, CECILL-1.1, CECILL-2.0, <br> CECILL-2.1,  CECILL-C, EPL-1.0, EPL-2.0, <br> LGPL-2.1, LGPL-3.0, MS-RL, MPL-2.0     | Weak Copyleft |
@@ -563,6 +563,15 @@ Instead of removing exact and near duplicates found against other open datasets,
 The final dataset structure is shown in the example below.
 ```json
 {
     "file_name": "Font.java",
@@ -606,6 +615,67 @@ The final dataset structure is shown in the example below.
 - **exact_duplicates_pubdataset**: boolean flag stating if there are any exact duplicate files found against another public dataset (The Stackv2, The Stack, RedPajama, GithubCode, CodeParrot)
 - **near_duplicates_pubdataset**: boolean flag stating if there are any near duplicate files found against another public dataset (The Stackv2, The Stack, RedPajama, GithubCode, CodeParrot)
 # Usage

 We collect up to **50,000** public repositories using the [GitHub API](https://docs.github.com/en/rest/search/search?apiVersion=2022-11-28), focusing on *license type*, *star count*, and *creation date*. Repositories with non-permissive licenses are prioritized to reduce contamination, as public code datasets we deduplicate against primarily focus on permissive or no-license repositories. We select repositories created before **April 2024** in decreasing order of their star counts. To handle GitHub rate limits, we use timeouts and pagination during the scraping process.
+### Copyleft licenses included in the The Heap
 | **License**    | **Family** |
 | :--------: | :-------: |
 | CECILL-1.0, CECILL-1.1, CECILL-2.0, <br> CECILL-2.1,  CECILL-C, EPL-1.0, EPL-2.0, <br> LGPL-2.1, LGPL-3.0, MS-RL, MPL-2.0     | Weak Copyleft |
 The final dataset structure is shown in the example below.
+### Open code dataset used for deduplication
+| **Dataset**    | **Size** |
+| :--------: | :-------: |
+| The Stack V2  | 67.5 TB   |
+| The Stack   | 6.4 TB    |
+| Red Pajama |    2.67 TB    |
+| GitHub Code |    1 TB    |
+| CodeParrot | 180 GB |
 ```json
 {
     "file_name": "Font.java",
 - **exact_duplicates_pubdataset**: boolean flag stating if there are any exact duplicate files found against another public dataset (The Stackv2, The Stack, RedPajama, GithubCode, CodeParrot)
 - **near_duplicates_pubdataset**: boolean flag stating if there are any near duplicate files found against another public dataset (The Stackv2, The Stack, RedPajama, GithubCode, CodeParrot)
+The distribution of the languages in The Heap is presented in the table below.
+### Programming languages included in The Heap
+| **Language** | **Repositories** | **Raw Files** | **Unique Files** |
+| --- | --- | --- | --- |
+| Ada | 676 | 41,367 | 35,425 |
+| Agda | 142 | 5,483 | 5,113 |
+| ANTLR | 101 | 564 | 541 |
+| Apex | 254 | 17,833 | 13,641 |
+| Assembly | 7,100 | 208,896 | 104,911 |
+| Awk | 1,191 | 16,586 | 15,620 |
+| C# | 50,000 | 5,906,716 | 3,770,829 |
+| C++ | 50,000 | 14,891,856 | 8,341,620 |
+| Clojure | 27,107 | 380,567 | 273,181 |
+| COBOL | 241 | 2,242 | 1,208 |
+| Common Lisp | 796 | 45,083 | 16,968 |
+| Cuda | 11,907 | 54,137 | 26,175 |
+| Crystal | 368 | 11,606 | 7,300 |
+| D | 1,191 | 26,048 | 13,359 |
+| Dart | 1,185 | 185,630 | 128,111 |
+| Elixir | 11,907 | 484,935 | 413,203 |
+| Elm | 1,477 | 15,511 | 12,384 |
+| Erlang | 2,475 | 453,856 | 127,910 |
+| F# | 277 | 8,260 | 7,963 |
+| Forth | 1,240 | 55,932 | 32,049 |
+| Fortran | 876 | 22,152 | 16,015 |
+| Groovy | 222 | 28,287 | 7,932 |
+| Hack | 2,198 | 60,299 | 48,353 |
+| Haskell | 1,279 | 84,916 | 37,405 |
+| JavaScript | 8,023 | 122,788 | 111,234 |
+| Julia | 50,000 | 6,989,601 | 3,757,338 |
+| Kotlin | 2,959 | 46,284 | 38,381 |
+| Less | 21,665 | 1,467,343 | 1,389 |
+| Lisp | 433 | 17,276 | 7,389 |
+| Lua | 42,241 | 4,605,230 | 912,898 |
+| Mathematica | 1,528 | 164,498 | 89,853 |
+| MATLAB | 20,828 | 1,051,354 | 665,659 |
+| NetLogo | 332 | 900 | 863 |
+| NewLisp | 35 | 5,819 | 5,148 |
+| Nix | 1,892 | 75,093 | 71,199 |
+| Objective-C | 7,700 | 1,899,714 | 698,137 |
+| OCaml | 1,761 | 321,989 | 95,171 |
+| Pascal | 5,218 | 130,832 | 225,749 |
+| Perl | 50,000 | 1,798,520 | 269,760 |
+| PHP | 50,000 | 12,707,727 | 3,363,040 |
+| Processing | 2,950 | 24,723 | 20,343 |
+| Prolog | 1,071 | 38,995 | 20,279 |
+| Python | 50,000 | 2,290,182 | 1,792,451 |
+| R | 44,993 | 389,139 | 374,812 |
+| Racket | 158 | 1,384 | 1,306 |
+| Ruby | 13,378 | 1,579,635 | 794,364 |
+| Rust | 42,847 | 2,496,177 | 844,258 |
+| Scala | 5,893 | 749,370 | 224,021 |
+| Scheme | 1,878 | 106,620 | 53,226 |
+| Shell | 150 | 47,531 | 1,084 |
+| SQL | 130 | 47,185 | 41,178 |
+| Swift | 13,924 | 633,819 | 439,565 |
+| Vue | 14,858 | 834 | 498 |
+| WebAssembly | 68 | 834 | 587 |
+| Total | 733,663 | 96,990,250 | 38,681,690 |
 # Usage