stats.sh: process files in memory instead of on disk

Writing the file to the filesystem for all tags is unnecessary if the
content can just be parsed in memory. This improves the performance:

Here the output of `time ./stats.sh`:

old
./stats.sh  5.96s user 26.53s system 94% cpu 34.533 total

new
./stats.sh  5.85s user 22.86s system 102% cpu 28.040 total

So a reduction of 6.5s (23%) is achieved even on a computer with a
decent SSD.

diff --git a/stats.sh b/stats.sh
index 2053501fdbf570367a08aa3eef60be7fed1f513f..ff4dabea0a4b41f54fef67e4be17fa9b5bca70c4 100755 (executable)
--- a/stats.sh
+++ b/stats.sh
@@ -1,12 +1,13 @@
 #!/usr/bin/env bash
 
-echo \n "" > stats.out
+# clear file
+true > stats.out
 
 for TAG_DATE in $(git tag --sort=creatordate  --format='%(refname:short),%(creatordate:short)'); do
   # echo "$TAG_DATE"
   split=(${TAG_DATE//,/ })
   # echo ${split[0]}
-  git checkout tags/${split[0]} readmeData.json
-  entries=$(jq '.base.entries' readmeData.json)
+  entries=$(git show tags/${split[0]}:readmeData.json | jq '.base.entries')
+  if [[ -z "$entries" ]]; then continue; fi
   echo ${split[1]},${entries} >> stats.out
 done