1-- Generate n-grams of Skia API calls from SKPs. 2 3-- To test this locally, run: 4-- $ GYP_DEFINES="skia_shared_lib=1" make lua_pictures 5-- $ out/Debug/lua_pictures -q -r $SKP_DIR -l tools/lua/ngrams.lua > /tmp/lua-output 6-- $ lua tools/lua/ngrams_aggregate.lua 7 8-- To run on Cluster Telemetry, copy and paste the contents of this file into 9-- the box at https://skia-tree-status.appspot.com/skia-telemetry/lua_script, 10-- and paste the contents of ngrams_aggregate.lua into the "aggregator script" 11-- box on the same page. 12 13-- Change n as desired. 14-- CHANGEME 15local n = 3 16-- CHANGEME 17 18-- This algorithm uses a list-of-lists for each SKP. For API call, append a 19-- list containing just the verb to the master list. Then, backtrack over the 20-- last (n-1) sublists in the master list and append the verb to those 21-- sublists. At the end of execution, the master list contains a sublist for 22-- every verb in the SKP file. Each sublist has length n, with the exception of 23-- the last n-1 sublists, which are discarded in the summarize() function, 24-- which generates counts for each n-gram. 25 26local ngrams = {} 27local currentFile = "" 28 29function sk_scrape_startcanvas(c, fileName) 30 currentFile = fileName 31 ngrams[currentFile] = {} 32end 33 34function sk_scrape_endcanvas(c, fileName) 35end 36 37function sk_scrape_accumulate(t) 38 table.insert(ngrams[currentFile], {t.verb}) 39 for i = 1, n-1 do 40 local idx = #ngrams[currentFile] - i 41 if idx > 0 then 42 table.insert(ngrams[currentFile][idx], t.verb) 43 end 44 end 45end 46 47function sk_scrape_summarize() 48 -- Count the n-grams. 49 local counts = {} 50 for file, ngramsInFile in pairs(ngrams) do 51 for i = 1, #ngramsInFile - (n-1) do 52 local ngram = table.concat(ngramsInFile[i], " ") 53 if counts[ngram] == nil then 54 counts[ngram] = 1 55 else 56 counts[ngram] = counts[ngram] + 1 57 end 58 end 59 end 60 61 -- Write out code for aggregating. 62 for ngram, count in pairs(counts) do 63 io.write("if counts['", ngram, "'] == nil then counts['", ngram, "'] = ", count, " else counts['", ngram, "'] = counts['", ngram, "'] + ", count, " end\n") 64 end 65end 66