arbitrary attributes: fixes and bringing lua up to date

Author: bottler

results/Readme.md

@@ -16,6 +16,9 @@ This stores the reported training and test objective and accuracy from each step
 These are stored in the STEPS table.
 The time for the first 10 steps (excluding the first) is remembered in the steps table.
 
+#### `results.addResultAttribute`
+Stores a value of an attribute for the run, which must not have a value already.
+
 Similar functions are also available in lua after something like this.
 ```
 results = require 'results'

results/results.lua

@@ -6,6 +6,7 @@ local results = {}
 filename = os.getenv("JR_RESULTSFILE") or "results.sqlite"
 con = sqlite3.open(filename)
 filecontents = ""
+oldColumns=False
 
 function append_file_to_filecontents_string(f)
    local ff = io.open(f,"r")
@@ -24,43 +25,93 @@ end
 --resultless(sql, param1, param2) executes it with the given parameters
 function resultless(st, ...)
    local s = con:prepare(st)
-   assert(s)
+   if not s then
+      error(con:errmsg())
+   end
    s:bind_values(...)
    local res = s:step()
+   if res==sqlite3.ROW then
+      error("resultless: the query returned something")
+   end
    if res~=sqlite3.DONE then
       error(con:errmsg())
    end
    s:finalize()
 end
+function singleValuedQuery(st, ...)
+   local s = con:prepare(st)
+   if not s then
+      error(con:errmsg())
+   end
+   s:bind_values(...)
+   local res = s:step()
+   if res==sqlite3.DONE then
+      error("the query returned nothing")
+   end
+   if res~=sqlite3.ROW then
+      error(con:errmsg())
+   end
+   if s:columns()~= 1 then
+      error("the query returned multiple columns")
+   end
+   local o = s:get_value(0)
+   res=s:step()
+   if res~=sqlite3.DONE then
+      error("perhaps the query returned more than one row")
+   end
+   s:finalize()
+   return o
+end
 
-function results.startRun(repnname, repnlength, continuation, batchTr, batchTe, layertype, layers, width, architecture, solver, callerFilePath)
+function results.startRun(continuation, attribs, architecture, solver, callerFilePath)
    if "table" == type(callerFilePath) then
       for i,name in ipairs(callerFilePath) do
          append_file_to_filecontents_string(name)
       end
    else
       append_file_to_filecontents_string(callerFilePath)
    end
-   local setup=[[create table if not exists RUNS(COUNT INTEGER PRIMARY KEY, TIME TEXT DEFAULT CURRENT_TIMESTAMP NOT NULL, REPN TEXT, REPNLENGTH INT, CONTINUATION TEXT, BATCHTR INT, BATCHTE INT, LAYERTYPE TEXT, LAYERS INT, WIDTH INT, ARCHITECTURE TEXT, SOLVER TEXT, CODE TEXT)   ;
-   create table if not exists STEPS(STEP INTEGER PRIMARY KEY, RUN int, OBJECTIVE real, TRAINACC real, TESTOBJECTIVE real, TESTACC REAL );
-   create table if not exists TIMES(RUN INT, TIME real)
+   local runColList="(COUNT INTEGER PRIMARY KEY, TIME TEXT DEFAULT CURRENT_TIMESTAMP NOT NULL, REPN TEXT, REPNLENGTH INT, CONTINUATION TEXT, BATCHTR INT, BATCHTE INT, LAYERTYPE TEXT, LAYERS INT, WIDTH INT, ARCHITECTURE TEXT, SOLVER TEXT, CODE TEXT)"
+   if not oldColumns then
+      runColList="(COUNT INTEGER PRIMARY KEY, TIME TEXT DEFAULT CURRENT_TIMESTAMP NOT NULL, CONTINUATION TEXT, ARCHITECTURE TEXT, SOLVER TEXT, CODE TEXT)"
+   end
+   local setup="create table if not exists RUNS" .. runColList .. [[;
+      create table if not exists STEPS(STEP INTEGER PRIMARY KEY, RUN int, OBJECTIVE real, TRAINACC real, TESTOBJECTIVE real, TESTACC REAL );
+      create table if not exists TIMES(RUN INT, TIME real);
+      create table if not exists ATTRIBS(RUN INT, NAME TEXT, ISRESULT INT, VALUE TEXT)
    ]]
    check(con:exec(setup))
-   local infoquery = "insert into RUNS (REPN, REPNLENGTH, CONTINUATION, BATCHTR, BATCHTE, LAYERTYPE, LAYERS, WIDTH, ARCHITECTURE, SOLVER, CODE) VALUES (?,?,?,?,?,?,?,?,?,?,?)"
-   local info = {repnname, repnlength, continuation, batchTr, batchTe, layertype, layers, width, architecture, solver, filecontents}
+   local infoquery = "insert into RUNS (CONTINUATION, ARCHITECTURE, SOLVER, CODE) VALUES (?,?,?,?)"
+   local info = {continuation, architecture, solver, filecontents}
+   local attribquery = "insert into ATTRIBS(RUN, NAME, ISRESULT, VALUE) VALUES (?,?,0,?)"
    resultless(infoquery,unpack(info))
    nrun = con:last_insert_rowid()
-   sys.tic()
+   for k,v in pairs(attribs) do
+      resultless(attribquery,nrun,k,v)
+   end
    nsteps = 0
    filecontents = nil
 end
 
 function results.step(obj,train,objte,test)
    nsteps = 1+nsteps
    resultless("insert into steps values (NULL, ?, ?, ?, ?, ?)", nrun,obj,train,objte, test)
-   if nsteps == 10 then
+   if nsteps == 1 then
+      sys.tic()
+   end
+   if nsteps == 11 then
       resultless("insert into TIMES VALUES (?,?)", nrun, sys.toc())
    end
 end
 
+function results.addResultAttribute(name, value)
+   if 1 then
+      count = singleValuedQuery("select count(*) from ATTRIBS where run = ? and name = ?",nrun,name)
+      if count > 0 then
+         error("attribute already set:" .. name)
+      end
+   end
+   resultless("insert into ATTRIBS(RUN, NAME, ISRESULT, VALUE) VALUES (?,?,1,?)",nrun,name,value)
+end
+
 return results

results/results.py

@@ -157,9 +157,12 @@ def runList():
     print ( tabulate.tabulate([i for i in b],headers=["","","steps", "start","obj","trainAcc","testObj","testAcc","Bad"]) )
     #return a,b
 
+def _getUseOldColumns():
+    return oldColumns and c.execute("select sql like '%LAYERTYPE%' from sqlite_master where tbl_name='RUNS'").fetchone()[0]
+    
 def runList1():
     c=con.cursor()
-    if not oldColumns:
+    if not _getUseOldColumns():
         raise RuntimeError("runList1 not available in the new format")
     b=c.execute("select repnlength, repn, count(step), min(objective), max(trainacc), max(testacc) from runs r left join steps s on r.count=s.run group by count").fetchall()
     for j in b:
@@ -192,8 +195,9 @@ def runList2(avgLength=None, doPrint = True, runFrom=None, architectureLike=None
     b=[]
     archClause = ""
     masterQueryArgs = None
+    useOldColumns=_getUseOldColumns()
     if architectureLike is not None:
-        if not oldColumns:
+        if not useOldColumns:
             raise RuntimeError("no architecture search with new format")
 #        architectureStr = c.execute("select architecture from runs where count = ?",(architectureLike,)).fetchone()[0]
 #        archClause = "where architecture = ?"
@@ -203,7 +207,7 @@ def runList2(avgLength=None, doPrint = True, runFrom=None, architectureLike=None
         if runFrom is not None:
             archClause = archClause + " and count >= " + str(runFrom)
             runFrom = None
-    masterQuery = "select count, "+("repnlength, repn," if oldColumns else "") +" case when length(continuation)>0 then '+' else '' end || count(step), (select time from times where run = r.count) from runs r left join steps s on r.count=s.run %s group by count" % (archClause if runFrom is None else ("where count>= "+str(runFrom)))
+    masterQuery = "select count, "+("repnlength, repn," if useOldColumns else "") +" case when length(continuation)>0 then '+' else '' end || count(step), (select time from times where run = r.count) from runs r left join steps s on r.count=s.run %s group by count" % (archClause if runFrom is None else ("where count>= "+str(runFrom)))
     if masterQueryArgs is None:
         bb=c.execute(masterQuery).fetchall()
     else:
@@ -217,9 +221,9 @@ def bestAvg(idx,highIsBad):
             s=replaceSeriesWithMovingAverage_([i[idx] for i in values],avgLength,highIsBad)
             return (numpy.amin if highIsBad else numpy.amax)(s)
         attribValues=tuple(_getAttribAsStr(rec[0],att) for att in attribs)
-        b.append(rec[:4] + (bestAvg(0,True),bestAvg(1,False),bestAvg(2,False))+attribValues)
+        b.append(rec[:-1] + (bestAvg(0,True),bestAvg(1,False),bestAvg(2,False),rec[-1])+attribValues)
     if doPrint:
-        if oldColumns:
+        if useOldColumns:
             headers=["","repLen","repn","steps","objective","trainAcc","testAcc","10StepTime"]+attribs
         else:
             headers=["",                "steps","objective","trainAcc","testAcc","10StepTime"]+attribs