Add script to compute the number of bytes per file type

kgryte · kgryte · commit 3d5d4c056ae0 · 2016-10-20T18:36:13.000-07:00
diff --git a/tools/git/scripts/bytes_per_file_type b/tools/git/scripts/bytes_per_file_type
@@ -0,0 +1,216 @@
+#!/usr/bin/env bash
+#
+# Prints an approximate number of bytes per file type.
+#
+# <proportion> <total_bytes> <number_of_files> <average> <file_type>
+#
+# Notes:
+#
+# * `README.md` files are included in Markdown statistics.
+
+# * `git ls-files`
+#   - List indexed files.
+# * `ls -l`
+#   - Print file info, which includes number of bytes.
+# * `awk '{}'`
+#   - Compute statistics.
+# * `sort -nr`
+#   - Sort in reverse numeric order.
+git ls-files | xargs ls -l | awk '
+{
+	bytes["total"] += $5
+	N["total"] += 1
+}
+
+# Special cases...
+/(LICENSE|NOTICE)$/ {
+	bytes["LICENSE"] += $5
+	N["LICENSE"] += 1
+	next
+}
+/datapackage\.json$/ {
+	bytes["datapackage.json"] += $5
+	N["datapackage.json"] += 1
+	next
+}
+/package\.json$/ {
+	bytes["package.json"] += $5
+	N["package.json"] += 1
+	next
+}
+
+# Known file extensions (keep in alphabetical order)...
+/\.awk$/ {
+	bytes["AWK"] += $5
+	N["AWK"] += 1
+	next
+}
+/\.bib$/ {
+	bytes["BibTeX"] += $5
+	N["BibTeX"] += 1
+	next
+}
+/\.c$/ {
+	bytes["C"] += $5
+	N["C"] += 1
+	next
+}
+/\.cpp$/ {
+	bytes["C++"] += $5
+	N["C++"] += 1
+	next
+}
+/\.csl$/ {
+	bytes["CSL"] += $5
+	N["CSL"] += 1
+	next
+}
+/\.css$/ {
+	bytes["CSS"] += $5
+	N["CSS"] += 1
+	next
+}
+/\.csv$/ {
+	bytes["CSV"] += $5
+	N["CSV"] += 1
+	next
+}
+/\.eot$/ {
+	bytes["fonts"] += $5
+	N["fonts"] += 1
+	next
+}
+/\.gif$/ {
+	bytes["gif"] += $5
+	N["gif"] += 1
+	next
+}
+/\.go$/ {
+	bytes["Go"] += $5
+	N["Go"] += 1
+	next
+}
+/\.h$/ {
+	bytes["C"] += $5
+	N["C"] += 1
+	next
+}
+/\.hpp$/ {
+	bytes["C++"] += $5
+	N["C++"] += 1
+	next
+}
+/\.html$/ {
+	bytes["HTML"] += $5
+	N["HTML"] += 1
+	next
+}
+/\.jl$/ {
+	bytes["Julia"] += $5
+	N["Julia"] += 1
+	next
+}
+/\.jpg$/ {
+	bytes["JPG"] += $5
+	N["JPG"] += 1
+	next
+}
+/\.js$/ {
+	bytes["JavaScript"] += $5
+	N["JavaScript"] += 1
+	next
+}
+/\.json$/ {
+	bytes["JSON"] += $5
+	N["JSON"] += 1
+	next
+}
+/Makefile$/ {
+	bytes["make"] += $5
+	N["make"] += 1
+	next
+}
+/\.md$/ {
+	if ($9 ~/README\.md$/) {
+		bytes["README"] += $5
+		N["README"] += 1
+	}
+	bytes["Markdown"] += $5
+	N["Markdown"] += 1
+	next
+}
+/\.mk$/ {
+	bytes["make"] += $5
+	N["make"] += 1
+	next
+}
+/\.png$/ {
+	bytes["PNG"] += $5
+	N["PNG"] += 1
+	next
+}
+/\.py$/ {
+	bytes["Python"] += $5
+	N["Python"] += 1
+	next
+}
+/\.R$/ {
+	bytes["R"] += $5
+	N["R"] += 1
+	next
+}
+/\.sh$/ {
+	bytes["bash"] += $5
+	N["bash"] += 1
+	next
+}
+/\.svg$/ {
+	bytes["SVG"] += $5
+	N["SVG"] += 1
+	next
+}
+/\.txt$/ {
+	bytes["plaintext"] += $5
+	N["plaintext"] += 1
+	next
+}
+/\.woff$/ {
+	bytes["fonts"] += $5
+	N["fonts"] += 1
+	next
+}
+/\.yml$/ {
+	bytes["YAML"] += $5
+	N["YAML"] += 1
+	next
+}
+
+# Special cases...
+$9 ~ /^\.([A-Za-z])+$|\/\.([A-Za-z])+$/ {
+	bytes["dotfiles"] += $5
+	N["dotfiles"] += 1
+	next
+}
+$9 ~ /\/([A-Za-z0-9_-])+$/ {
+	bytes["executables"] += $5
+	N["executables"] += 1
+	next
+}
+
+# Everything else...
+{
+	bytes["other"] += $5
+	N["other"] += 1
+}
+
+END {
+	total = bytes["total"]
+	for (k in bytes) {
+		len = bytes[k];
+		n = N[k]
+		pct = int(len/total*10000)/10000
+		mu = int(len/n*10000)/10000
+		print pct OFS len OFS n OFS mu OFS k
+	}
+}
+' | sort -nr
