added blog post

_posts/2024-05-24-self-assembly.html

@@ -0,0 +1,67 @@
+---
+title: Aider has written 7% of its own code
+excerpt: Aider is 14K lines of python and 7% of those were written by aider itself.
+highlight_image: /assets/self-assembly.jpg
+draft: true
+---
+
+# Aider has written 7% of its own code
+
+[![SWE Bench Lite results](/assets/self-assembly.jpg)](https://aider.chat/assets/self-assembly.jpg)
+
+The
+[aider git repo]
+currently contains about 4K commits, about 15% of which
+were made by aider.
+
+The code is currently about 14K lines of python.
+Aider wrote about 7% of those lines of code.
+In the history of the repo, aider has added 4.8K lines of code
+and deleted 1.5K lines.
+
+Here's the breakdown of the code aider wrote in the current code base.
+
+| File | Lines | Percent |
+|---|---:|---:|
+|aider/args.py| 6 of 449 | 1.34% |
+|aider/coders/base_coder.py| 37 of 1354 | 2.73% |
+|aider/coders/editblock_coder.py| 10 of 507 | 1.97% |
+|aider/coders/editblock_func_coder.py| 8 of 141 | 5.67% |
+|aider/coders/udiff_coder.py| 2 of 421 | 0.48% |
+|aider/coders/wholefile_coder.py| 5 of 146 | 3.42% |
+|aider/coders/wholefile_func_coder.py| 11 of 134 | 8.21% |
+|aider/commands.py| 68 of 703 | 9.67% |
+|aider/diffs.py| 15 of 129 | 11.63% |
+|aider/gui.py| 2 of 533 | 0.38% |
+|aider/history.py| 18 of 124 | 14.52% |
+|aider/io.py| 45 of 368 | 12.23% |
+|aider/linter.py| 29 of 240 | 12.08% |
+|aider/main.py| 26 of 466 | 5.58% |
+|aider/mdstream.py| 2 of 122 | 1.64% |
+|aider/models.py| 18 of 549 | 3.28% |
+|aider/repo.py| 20 of 266 | 7.52% |
+|aider/repomap.py| 17 of 518 | 3.28% |
+|aider/scrape.py| 11 of 199 | 5.53% |
+|aider/versioncheck.py| 10 of 37 | 27.03% |
+|aider/voice.py| 8 of 104 | 7.69% |
+|benchmark/benchmark.py| 32 of 730 | 4.38% |
+|benchmark/over_time.py| 32 of 60 | 53.33% |
+|benchmark/swe_bench_lite.py| 39 of 71 | 54.93% |
+|scripts/blame.py| 65 of 214 | 30.37% |
+|scripts/versionbump.py| 92 of 123 | 74.80% |
+|setup.py| 11 of 47 | 23.40% |
+|tests/test_coder.py| 40 of 612 | 6.54% |
+|tests/test_commands.py| 138 of 588 | 23.47% |
+|tests/test_editblock.py| 23 of 403 | 5.71% |
+|tests/test_io.py| 30 of 65 | 46.15% |
+|tests/test_main.py| 25 of 239 | 10.46% |
+|tests/test_models.py| 6 of 28 | 21.43% |
+|tests/test_repo.py| 2 of 296 | 0.68% |
+|tests/test_repomap.py| 70 of 217 | 32.26% |
+|tests/test_udiff.py| 7 of 119 | 5.88% |
+|tests/test_wholefile.py| 23 of 321 | 7.17% |
+|**TOTAL**| **1003 of 14221** | **7.05%** |
+
+
+
+

scripts/blame.py

@@ -1,32 +1,32 @@
 #!/usr/bin/env python3
 
-import tempfile
-import sys
 import subprocess
+import sys
+import tempfile
 from pathlib import Path
+
 import pylab as plt
-from aider.dump import dump
 from imgcat import imgcat
 
+from aider.dump import dump
+
+
 def get_lines_with_commit_hash(filename, aider_commits, git_dname, verbose=False):
     result = subprocess.run(
         ["git", "-C", git_dname, "blame", "-l", filename],
         capture_output=True,
         text=True,
-        check=True
+        check=True,
     )
 
-    hashes = [
-        line.split()[0]
-        for line in result.stdout.splitlines()
-    ]
+    hashes = [line.split()[0] for line in result.stdout.splitlines()]
     lines = Path(filename).read_text().splitlines()
 
     num_aider_lines = 0
-    for hsh,line in zip(hashes, lines):
+    for hsh, line in zip(hashes, lines):
         if hsh in aider_commits:
             num_aider_lines += 1
-            prefix = '+'
+            prefix = "+"
         else:
             prefix = " "
 
@@ -45,7 +45,7 @@ def get_aider_commits(git_dname):
         ["git", "-C", git_dname, "log", "--pretty=format:%H %s"],
         capture_output=True,
         text=True,
-        check=True
+        check=True,
     )
 
     results = result.stdout.splitlines()
@@ -59,7 +59,6 @@ def get_aider_commits(git_dname):
 
     dump(len(commits))
 
-
     return commits
 
 
@@ -72,17 +71,28 @@ def show_commit_stats(commits):
             ["git", "show", "--stat", "--oneline", commit],
             capture_output=True,
             text=True,
-            check=True
+            check=True,
         )
 
         added_lines = 0
         deleted_lines = 0
         for line in result.stdout.splitlines():
+            if "changed," not in line:
+                continue
+            if "insertion" not in line and "deletion" not in line:
+                continue
             dump(line)
-            if "insertion" in line or "insertion(+)" in line:
-                added_lines += int(line.split()[line.split().index("insertion(+)") - 1])
-            if "deletion" in line or "deletion(-)" in line:
-                deleted_lines += int(line.split()[line.split().index("deletion(-)") - 1])
+            pieces = line.split(",")
+            try:
+                for piece in pieces:
+                    if "insertion" in piece:
+                        dump(piece)
+                        added_lines += int(piece.strip().split()[0])
+                    if "deletion" in piece:
+                        dump(piece)
+                        deleted_lines += int(piece.strip().split()[0])
+            except ValueError:
+                pass
 
         total_added_lines += added_lines
         total_deleted_lines += deleted_lines
@@ -91,12 +101,13 @@ def show_commit_stats(commits):
 
     print(f"Total: +{total_added_lines} -{total_deleted_lines}")
 
+
 def process_fnames(fnames, git_dname):
     if not git_dname:
         git_dname = "."
 
     aider_commits = get_aider_commits(git_dname)
-    show_commit_stats(aider_commits)
+    # show_commit_stats(aider_commits)
 
     total_lines = 0
     total_aider_lines = 0
@@ -108,24 +119,24 @@ def process_fnames(fnames, git_dname):
         percent_modified = (num_aider_lines / num_lines) * 100 if num_lines > 0 else 0
         if not num_aider_lines:
             continue
-        print(f"{fname}: {num_aider_lines}/{num_lines} ({percent_modified:.2f}%)")
+        print(f"|{fname}| {num_aider_lines} of {num_lines} | {percent_modified:.2f}% |")
 
     total_percent_modified = (total_aider_lines / total_lines) * 100 if total_lines > 0 else 0
-    print(f"Total: {total_aider_lines}/{total_lines} lines by aider ({total_percent_modified:.2f}%)")
+    print(
+        f"Total: {total_aider_lines}/{total_lines} lines by aider ({total_percent_modified:.2f}%)"
+    )
     return total_aider_lines, total_lines, total_percent_modified
 
+
 def process_repo(git_dname=None):
     if not git_dname:
         git_dname = "."
 
     result = subprocess.run(
-        ["git", "-C", git_dname, "ls-files"],
-        capture_output=True,
-        text=True,
-        check=True
+        ["git", "-C", git_dname, "ls-files"], capture_output=True, text=True, check=True
     )
     git_dname = Path(git_dname)
-    fnames = [git_dname/fname for fname in result.stdout.splitlines() if fname.endswith('.py')]
+    fnames = [git_dname / fname for fname in result.stdout.splitlines() if fname.endswith(".py")]
 
     return process_fnames(fnames, git_dname)
 
@@ -136,7 +147,7 @@ def history():
         ["git", "-C", git_dname, "log", "--pretty=format:%H %s"],
         capture_output=True,
         text=True,
-        check=True
+        check=True,
     )
 
     commits = []
@@ -148,8 +159,8 @@ def history():
     dump(len(commits))
 
     num_commits = len(commits)
-    N=10
-    step = (num_commits-1)/(N-1)
+    N = 10
+    step = (num_commits - 1) / (N - 1)
     results = []
     i = 0
     while i < num_commits:
@@ -176,21 +187,19 @@ def history():
     aider_lines = [aider_lines for _, aider_lines, _, _ in results]
     total_lines = [total_lines for _, _, total_lines, _ in results]
 
-    plt.fill_between(x, aider_lines, label='Aider Lines', color='skyblue', alpha=0.5)
-    plt.fill_between(x, total_lines, label='Total Lines', color='lightgreen', alpha=0.5)
-    plt.xlabel('Commit Number')
-    plt.ylabel('Lines of Code')
-    plt.title('Aider Lines and Total Lines Over Time')
+    plt.fill_between(x, aider_lines, label="Aider Lines", color="skyblue", alpha=0.5)
+    plt.fill_between(x, total_lines, label="Total Lines", color="lightgreen", alpha=0.5)
+    plt.xlabel("Commit Number")
+    plt.ylabel("Lines of Code")
+    plt.title("Aider Lines and Total Lines Over Time")
     plt.legend()
-    plt.savefig('aider_plot.png')
-    with open('aider_plot.png', 'rb') as f:
+    plt.savefig("aider_plot.png")
+    with open("aider_plot.png", "rb") as f:
         imgcat(f.read())
 
 
-
-
 def main():
-    #return history()
+    # return history()
 
     if len(sys.argv) < 2:
         return process_repo()