void doSummaryStatsBed(struct sqlConnection *conn)
/* Put up page showing summary stats for track that is in database
* or that is bed-format custom. */
{
struct bed *bedList = NULL;
struct region *regionList = getRegions(), *region;
char *regionName = getRegionName();
long long regionSize = 0, gapTotal = 0, realSize = 0;
long startTime, midTime, endTime;
long loadTime = 0, calcTime = 0, freeTime = 0;
struct covStats *itemCovList = NULL, *blockCovList = NULL, *cov;
int itemCount = 0;
struct hTableInfo *hti = getHti(database, curTable, conn);
int minScore = BIGNUM, maxScore = -BIGNUM;
long long sumScores = 0;
boolean hasBlocks = hti->hasBlocks;
boolean hasScore = (hti->scoreField[0] != 0);
int fieldCount;
htmlOpen("%s (%s) Summary Statistics", curTableLabel(), curTable);
for (region = regionList; region != NULL; region = region->next)
{
struct lm *lm = lmInit(64*1024);
startTime = clock1000();
bedList = cookedBedList(conn, curTable, region, lm, &fieldCount);
if (fieldCount < 12)
hasBlocks = FALSE;
if (fieldCount < 5)
hasScore = FALSE;
midTime = clock1000();
loadTime += midTime - startTime;
if (bedList != NULL)
{
itemCount += slCount(bedList);
regionSize += region->end - region->start;
cov = calcSpanOverRegion(region, bedList);
slAddHead(&itemCovList, cov);
if (hasBlocks)
{
cov = calcBlocksOverRegion(region, bedList);
slAddHead(&blockCovList, cov);
}
if (hti->scoreField[0] != 0)
{
struct bed *bed;
for (bed = bedList; bed != NULL; bed = bed->next)
{
int score = bed->score;
if (score < minScore) minScore = score;
if (score > maxScore) maxScore = score;
sumScores += score;
}
}
}
endTime = clock1000();
calcTime += endTime - midTime;
lmCleanup(&lm);
bedList = NULL;
freeTime += clock1000() - endTime;
}
regionSize = basesInRegion(regionList, 0);
gapTotal = gapsInRegion(conn, regionList, 0);
realSize = regionSize - gapTotal;
hTableStart();
startTime = clock1000();
numberStatRow("item count", itemCount);
if (itemCount > 0)
{
cov = covStatsSum(itemCovList);
percentStatRow("item bases", cov->basesCovered, realSize);
percentStatRow("item total", cov->sumBases, realSize);
numberStatRow("smallest item", cov->minBases);
numberStatRow("average item", round((double)cov->sumBases/cov->itemCount));
numberStatRow("biggest item", cov->maxBases);
}
if (hasBlocks && itemCount > 0)
{
cov = covStatsSum(blockCovList);
hPrintf("<TR><TD>block count</TD><TD ALIGN=RIGHT>");
printLongWithCommas(stdout, cov->itemCount);
hPrintf("</TD></TR>\n");
percentStatRow("block bases", cov->basesCovered, realSize);
percentStatRow("block total", cov->sumBases, realSize);
numberStatRow("smallest block", cov->minBases);
numberStatRow("average block", round((double)cov->sumBases/cov->itemCount));
numberStatRow("biggest block", cov->maxBases);
}
if (hasScore != 0 && itemCount > 0 && sumScores != 0)
{
numberStatRow("smallest score", minScore);
numberStatRow("average score", round((double)sumScores/itemCount));
numberStatRow("biggest score", maxScore);
}
hTableEnd();
/* Show region and time stats part of stats page. */
webNewSection("Region and Timing Statistics");
hTableStart();
stringStatRow("region", regionName);
numberStatRow("bases in region", regionSize);
numberStatRow("bases in gaps", gapTotal);
floatStatRow("load time", 0.001*loadTime);
floatStatRow("calculation time", 0.001*calcTime);
floatStatRow("free memory time", 0.001*freeTime);
stringStatRow("filter", (anyFilter() ? "on" : "off"));
stringStatRow("intersection", (anyIntersection() ? "on" : "off"));
hTableEnd();
covStatsFreeList(&itemCovList);
covStatsFreeList(&blockCovList);
htmlClose();
}
void doSummaryStatsWiggle(struct sqlConnection *conn)
/* Put up page showing summary stats for wiggle track. */
{
// grab the right trackDb for the current table. The curTrack variable
// has the composite trackDb in it
struct trackDb *track = hTrackDbForTrack(database, curTable);
char *table = curTable;
struct region *region, *regionList = getRegions();
char *regionName = getRegionName();
long long regionSize = 0;
long long gapTotal = 0;
long startTime = 0, wigFetchTime = 0;
char splitTableOrFileName[HDB_MAX_TABLE_STRING];
struct customTrack *ct = NULL;
boolean isCustom = FALSE;
struct wiggleDataStream *wds = NULL;
unsigned long long valuesMatched = 0;
int regionCount = 0;
int regionsDone = 0;
unsigned span = 0;
char *dataConstraint;
double ll = 0.0;
double ul = 0.0;
boolean hasConstraint = FALSE;
char *table2 = NULL;
boolean fullGenome = FALSE;
boolean statsHeaderDone = FALSE;
boolean gotSome = FALSE;
char *shortLabel = table;
long long statsItemCount = 0; /* global accumulators for overall */
int statsSpan = 0; /* stats summary on a multiple region */
double statsSumData = 0.0; /* output */
double statsSumSquares = 0.0; /* " " */
double lowerLimit = INFINITY; /* " " */
double upperLimit = -1.0 * INFINITY; /* " " */
startTime = clock1000();
if (track != NULL)
shortLabel = track->shortLabel;
/* Count the regions, when only one, we can do more stats */
for (region = regionList; region != NULL; region = region->next)
++regionCount;
htmlOpen("%s (%s) Wiggle Summary Statistics", shortLabel, table);
if (anySubtrackMerge(database, curTable))
hPrintf("<P><EM><B>Note:</B> subtrack merge is currently ignored on this "
"page (not implemented yet). Statistics shown here are only for "
"the primary table %s (%s).</EM>", shortLabel, table);
fullGenome = fullGenomeRegion();
WIG_INIT; /* ct, isCustom, hasConstraint, wds and table2 are set here */
for (region = regionList; region != NULL; region = region->next)
{
struct bed *intersectBedList = NULL;
int operations;
++regionsDone;
if (table2)
intersectBedList = bedTable2(conn, region, table2);
operations = wigFetchStats;
#if defined(NOT)
/* can't do the histogram now, that operation times out */
if (1 == regionCount)
operations |= wigFetchAscii;
#endif
wds->setChromConstraint(wds, region->chrom);
if (fullGenome)
wds->setPositionConstraint(wds, 0, 0);
else
wds->setPositionConstraint(wds, region->start, region->end);
if (hasConstraint)
wds->setDataConstraint(wds, dataConstraint, ll, ul);
/* depending on what is coming in on regionList, we may need to be
* smart about how often we call getData for these custom tracks
* since that is potentially a large file read each time.
*/
if (isCustom)
{
if (ct->dbTrack)
{
struct sqlConnection *trashConn = hAllocConn(CUSTOM_TRASH);
struct trackDb *tdb = findTdbForTable(database, curTrack, table, ctLookupName);
span = minSpan(trashConn, splitTableOrFileName, region->chrom,
region->start, region->end, cart, tdb);
wds->setSpanConstraint(wds, span);
valuesMatched = getWigglePossibleIntersection(wds, region,
CUSTOM_TRASH, table2, &intersectBedList,
splitTableOrFileName, operations);
hFreeConn(&trashConn);
}
else
{
valuesMatched = getWigglePossibleIntersection(wds, region, NULL,
table2, &intersectBedList, splitTableOrFileName, operations);
/* XXX We need to properly get the smallest span for custom tracks */
/* This is not necessarily the correct answer here */
if (wds->stats)
span = wds->stats->span;
else
span = 1;
}
}
else
{
if (hFindSplitTable(database, region->chrom, table, splitTableOrFileName, sizeof splitTableOrFileName, NULL))
{
span = minSpan(conn, splitTableOrFileName, region->chrom,
region->start, region->end, cart, track);
wds->setSpanConstraint(wds, span);
valuesMatched = getWigglePossibleIntersection(wds, region,
database, table2, &intersectBedList, splitTableOrFileName,
operations);
if (intersectBedList)
span = 1;
}
}
/* when doing multiple regions, we need to print out each result as
* it happens to keep the connection open to the browser and
* prevent any timeout since this could take a while.
* (worst case test is quality track on panTro1)
*/
if (wds->stats)
statsItemCount += wds->stats->count;
if (wds->stats && (regionCount > 1) && (valuesMatched > 0))
{
double sumData = wds->stats->mean * wds->stats->count;
double sumSquares;
if (wds->stats->count > 1)
sumSquares = (wds->stats->variance * (wds->stats->count - 1)) +
((sumData * sumData)/wds->stats->count);
else
sumSquares = sumData * sumData;
/* global accumulators for overall summary */
statsSpan = wds->stats->span;
statsSumData += sumData;
statsSumSquares += sumSquares;
if (wds->stats->lowerLimit < lowerLimit)
lowerLimit = wds->stats->lowerLimit;
if ((wds->stats->lowerLimit + wds->stats->dataRange) > upperLimit)
upperLimit = wds->stats->lowerLimit + wds->stats->dataRange;
if (statsHeaderDone)
wds->statsOut(wds, database, "stdout", TRUE, TRUE, FALSE, TRUE);
else
{
wds->statsOut(wds, database, "stdout", TRUE, TRUE, TRUE, TRUE);
statsHeaderDone = TRUE;
}
wds->freeStats(wds);
gotSome = TRUE;
}
if ((regionCount > MAX_REGION_DISPLAY) &&
(regionsDone >= MAX_REGION_DISPLAY))
{
hPrintf("<TR><TH ALIGN=CENTER COLSPAN=12> Can not display more "
"than %d regions, <BR> would take too much time </TH></TR>\n",
MAX_REGION_DISPLAY);
break; /* exit this for loop */
}
} /*for (region = regionList; region != NULL; region = region->next) */
if (hasConstraint)
freeMem(dataConstraint); /* been cloned into wds */
if (1 == regionCount)
{
statsPreamble(wds, regionList->chrom, regionList->start, regionList->end,
span, valuesMatched, table2);
/* 3 X TRUE = sort results, html table output, with header,
* the FALSE means close the table after printing, no more rows to
* come. The case in the if() statement was already taken care of
* in the statsPreamble() printout. No need to do that again.
*/
if ( ! ((valuesMatched == 0) && table2) )
wds->statsOut(wds, database, "stdout", TRUE, TRUE, TRUE, FALSE);
regionSize = basesInRegion(regionList,0);
gapTotal = gapsInRegion(conn, regionList,0);
}
else
{ /* this is a bit of a kludge here since these printouts are done in the
* library source wigDataStream.c statsOut() function and
* this is a clean up of that. That function should be
* pulled out of there and made independent and more
* versatile.
*/
long long realSize;
double variance;
double stddev;
/* Too expensive to lookup the numbers for thousands of regions */
regionSize = basesInRegion(regionList,MAX_REGION_DISPLAY);
gapTotal = gapsInRegion(conn, regionList,MAX_REGION_DISPLAY);
realSize = regionSize - gapTotal;
/* close the table which was left open in the loop above */
if (!gotSome)
hPrintf("<TR><TH ALIGN=CENTER COLSPAN=12> No data found matching this request </TH></TR>\n");
hPrintf("<TR><TH ALIGN=LEFT> SUMMARY: </TH>\n");
hPrintf("\t<TD> </TD>\n"); /* chromStart */
hPrintf("\t<TD> </TD>\n"); /* chromEnd */
hPrintf("\t<TD ALIGN=RIGHT> ");
printLongWithCommas(stdout, statsItemCount);
hPrintf(" </TD>\n" );
hPrintf("\t<TD ALIGN=RIGHT> %d </TD>\n", statsSpan);
hPrintf("\t<TD ALIGN=RIGHT> ");
printLongWithCommas(stdout, statsItemCount*statsSpan);
hPrintf(" (%.2f%%) </TD>\n",
100.0*(double)(statsItemCount*statsSpan)/(double)realSize);
hPrintf("\t<TD ALIGN=RIGHT> %g </TD>\n", lowerLimit);
hPrintf("\t<TD ALIGN=RIGHT> %g </TD>\n", upperLimit);
hPrintf("\t<TD ALIGN=RIGHT> %g </TD>\n", upperLimit - lowerLimit);
if (statsItemCount > 0)
hPrintf("\t<TD ALIGN=RIGHT> %g </TD>\n", statsSumData/statsItemCount);
else
hPrintf("\t<TD ALIGN=RIGHT> 0.0 </TD>\n");
stddev = 0.0;
variance = 0.0;
if (statsItemCount > 1)
{
variance = (statsSumSquares -
((statsSumData * statsSumData)/(double) statsItemCount)) /
(double) (statsItemCount - 1);
if (variance > 0.0)
stddev = sqrt(variance);
}
hPrintf("\t<TD ALIGN=RIGHT> %g </TD>\n", variance);
hPrintf("\t<TD ALIGN=RIGHT> %g </TD>\n", stddev);
hPrintf("</TR>\n");
wigStatsTableHeading(stdout, TRUE);
hPrintf("</TABLE></TD></TR></TABLE></P>\n");
}
#if defined(NOT)
/* can't do the histogram now, that operation times out */
/* Single region, we can do the histogram */
if ((valuesMatched > 1) && (1 == regionCount))
{
float *valuesArray = NULL;
size_t valueCount = 0;
struct histoResult *histoGramResult;
/* convert the ascii data listings to one giant float array */
valuesArray = wds->asciiToDataArray(wds, valuesMatched, &valueCount);
/* histoGram() may return NULL if it doesn't work */
histoGramResult = histoGram(valuesArray, valueCount,
NAN, (unsigned) 0, NAN, (float) wds->stats->lowerLimit,
(float) (wds->stats->lowerLimit + wds->stats->dataRange),
(struct histoResult *)NULL);
printHistoGram(histoGramResult, TRUE); /* TRUE == html output */
freeHistoGram(&histoGramResult);
wds->freeAscii(wds);
wds->freeArray(wds);
}
#endif
wds->freeStats(wds);
wiggleDataStreamFree(&wds);
wigFetchTime = clock1000() - startTime;
webNewSection("Region and Timing Statistics");
hTableStart();
stringStatRow("region", regionName);
numberStatRow("bases in region", regionSize);
numberStatRow("bases in gaps", gapTotal);
floatStatRow("load and calc time", 0.001*wigFetchTime);
wigFilterStatRow(conn);
stringStatRow("intersection", cartUsualString(cart, hgtaIntersectTable, "off"));
hTableEnd();
htmlClose();
} /* void doSummaryStatsWiggle(struct sqlConnection *conn) */
void doSummaryStatsBigWig(struct sqlConnection *conn)
/* Put up page showing summary stats for bigWig track. */
{
struct trackDb *track = curTrack;
char *table = curTable;
char *shortLabel = (track == NULL ? table : track->shortLabel);
char *fileName = bigWigFileName(table, conn);
long startTime = clock1000();
htmlOpen("%s (%s) Big Wig Summary Statistics", shortLabel, table);
if (anySubtrackMerge(database, curTable))
hPrintf("<P><EM><B>Note:</B> subtrack merge is currently ignored on this "
"page (not implemented yet). Statistics shown here are only for "
"the primary table %s (%s).</EM>", shortLabel, table);
struct bbiFile *bwf = bigWigFileOpen(fileName);
struct region *region, *regionList = getRegions();
double sumData = 0, sumSquares = 0, minVal = 0, maxVal = 0;
bits64 validCount = 0;
if (!anyFilter() && !anyIntersection())
{
for (region = regionList; region != NULL; region = region->next)
{
struct bbiSummaryElement sum;
if (bbiSummaryArrayExtended(bwf, region->chrom, region->start, region->end,
bigWigIntervalQuery, 1, &sum))
{
if (validCount == 0)
{
minVal = sum.minVal;
maxVal = sum.maxVal;
}
else
{
if (sum.minVal < minVal)
minVal = sum.minVal;
if (sum.maxVal > maxVal)
maxVal = sum.maxVal;
}
sumData += sum.sumData;
sumSquares += sum.sumSquares;
validCount += sum.validCount;
}
}
}
else
{
double ll, ul;
enum wigCompare cmp;
getWigFilter(database, curTable, &cmp, &ll, &ul);
for (region = regionList; region != NULL; region = region->next)
{
struct lm *lm = lmInit(0);
struct bbiInterval *iv, *ivList;
ivList = intersectedFilteredBbiIntervalsOnRegion(conn, bwf, region, cmp, ll, ul, lm);
for (iv = ivList; iv != NULL; iv = iv->next)
{
double val = iv->val;
double size = iv->end - iv->start;
if (validCount == 0)
minVal = maxVal = val;
else
{
if (val < minVal)
minVal = val;
if (val > maxVal)
maxVal = val;
}
sumData += size*val;
sumSquares += size*val*val;
validCount += size;
}
lmCleanup(&lm);
}
}
hTableStart();
floatStatRow("mean", sumData/validCount);
floatStatRow("min", minVal);
floatStatRow("max", maxVal);
floatStatRow("standard deviation", calcStdFromSums(sumData, sumSquares, validCount));
numberStatRow("bases with data", validCount);
long long regionSize = basesInRegion(regionList,0);
long long gapTotal = gapsInRegion(conn, regionList,0);
numberStatRow("bases with sequence", regionSize - gapTotal);
numberStatRow("bases in region", regionSize);
wigFilterStatRow(conn);
stringStatRow("intersection", cartUsualString(cart, hgtaIntersectTable, "off"));
long wigFetchTime = clock1000() - startTime;
floatStatRow("load and calc time", 0.001*wigFetchTime);
hTableEnd();
bbiFileClose(&bwf);
htmlClose();
}
