This is pair of simple scripts for doing local full-text search among files of a given extension. Put them in ~/bin on linux (and probably mac too) and make them executable.

The search probably isn't best-practice (with a database of some sort of string prefices), but it's so fast for my common use of many small text/code files, and so simple to write and read, that I don't care.

Search results show line numbers and a little context around the result. It's not particularly extensible in itself, but short enough you could easily rewrite it to do something else, like return results in a more machine-parseable form.

I'll test here with Ubuntu 20.04, and files stored on an SSD. However, after find does its caching, it's no slower on another 16.04 machine with spinning rust.

Count the total number of files (for reference):

tsbertalan@sigurd:~$ cd ~/Dropbox/
tsbertalan@sigurd:~/Dropbox$ find . | wc
 513446 1329333 44711578

Half a million files.

Count the number of markdown files (for example):

tsbertalan@sigurd:~/Dropbox$ find . -iname "*.md" | wc
   2081    9562  162239

Actual usage here (remove the time at the beginning for normal, non-timed usage):

tsbertalan@sigurd:~/Dropbox$ time ffg md "This is a search test."
Searching for " This is a search test. " in md files.



===============================================================================
./testfile.md
1:This is a search test.
2-

real	0m1.302s
user	0m0.673s
sys	0m0.643s

Let's try a search with slightly more results:

tsbertalan@sigurd:~/Dropbox$ echo "ffg md the | wc" > slower_search.sh
tsbertalan@sigurd:~/Dropbox$ time bash slower_search.sh 
  67364  749145 7937814

real	0m8.722s
user	0m2.167s
sys	0m1.862s

How many actual results was that?

tsbertalan@sigurd:~/Dropbox$ ffg md the | grep ============= | wc
   1252    1252  100185

(This is admittedly a pretty rough way to count them, but good enough for government work.)

Let's do a clumsy search for function definitions in Python files:

tsbertalan@sigurd:~/Dropbox$ find . -iname "*.py" | wc
   8024   15736  624813
tsbertalan@sigurd:~/Dropbox$ echo "ffg py def | wc" > slower_search.sh
tsbertalan@sigurd:~/Dropbox$ time bash slower_search.sh 
 334322  999241 18839714

real	0m9.305s
user	0m6.502s
sys	0m2.617s
tsbertalan@sigurd:~/Dropbox$ ffg py def | grep ============= | wc
   5677    5721  454105

Hmm, only 5677 results among 8024 files, for a total of 334322 lines of search output.

...

After some testing, though I'm not really a C++ programmer, I find I somehow have more than 8000 cpp files in my Dropbox (mostly accounted for by some Arduino libraries, and redundant Eigen and Ipopt source trees). Let's search them for opening curly brackets.

tsbertalan@sigurd:~/Dropbox$ find . -iname "*.cpp" | wc
   8191   11799  758901
tsbertalan@sigurd:~/Dropbox$ echo "ffg cpp \"{\" | wc" > slower_search.sh
tsbertalan@sigurd:~/Dropbox$ cat slower_search.sh 
ffg cpp "{" | wc
tsbertalan@sigurd:~/Dropbox$ time bash slower_search.sh 
 519909 1767949 31093177

real	0m7.256s
user	0m6.075s
sys	0m2.098s
tsbertalan@sigurd:~/Dropbox$ ffg cpp "{" | grep ============= | wc
   5984    5999  478888

Still pretty fast. Times here are similar (12 seconds) if you instead search for the letter "e", since it really scales mostly with the number of files with the given extension it finds, not the number of query occurrences per file.