Analyzing a Codebase with JSON¶

The JSON export pairs well with jq. Export with the detail flags you need; each file then becomes an object carrying path, loc, and/or size, and each directory carries aggregate totals. The JSON format reference describes the full layout.

recursivist export \
  --format json --full-path \
  --exclude node_modules --exclude .git \
  --prefix structure --sort-by-loc --size

This writes structure.json. --full-path makes each path unique, which is what you want when listing files from several directories.

Totals¶

# Total lines of code
jq '.structure.loc // 0' structure.json

# Lines of code per top-level directory
jq -r '.structure.subdirectories // {} | to_entries[]
       | select(.value | has("loc"))
       | [.key, (.value.loc | tostring)] | @tsv' structure.json | sort -k2 -nr

Every File in the Tree¶

Files sit under files in each directory object, and directories nest under subdirectories, so listing every file means walking the tree. The remaining recipes share this jq prelude, which defines that walk and applies it to the root:

FILES='def files: (.files // [])[], ((.subdirectories // {})[] | files);
       .structure | files'

Largest Files¶

# Ten files with the most lines of code
jq -r "$FILES"' | select(has("loc")) | [.loc, .path] | @tsv' \
  structure.json | sort -nr | head -10

# Ten largest files by size
jq -r "$FILES"' | select(has("size")) | [.size, .path] | @tsv' \
  structure.json | sort -nr | head -10

By Extension¶

# Count files by extension
jq -r "$FILES"' | (.path | split(".") | .[-1]) | ascii_downcase' \
  structure.json | sort | uniq -c | sort -nr

# Lines of code by extension
jq -r "$FILES"' | select(has("loc"))
       | "\(.path | split(".") | .[-1])\t\(.loc)"' structure.json \
  | awk -F'\t' '{loc[$1]+=$2; n[$1]++}
      END {for (e in loc) printf "%-6s %8d lines in %d files\n", e, loc[e], n[e]}' \
  | sort -k2 -nr