diff options
Diffstat (limited to 'wqflask/utility/webqtlUtil.py')
-rwxr-xr-x | wqflask/utility/webqtlUtil.py | 981 |
1 files changed, 981 insertions, 0 deletions
diff --git a/wqflask/utility/webqtlUtil.py b/wqflask/utility/webqtlUtil.py new file mode 100755 index 00000000..99451e33 --- /dev/null +++ b/wqflask/utility/webqtlUtil.py @@ -0,0 +1,981 @@ +# Copyright (C) University of Tennessee Health Science Center, Memphis, TN. +# +# This program is free software: you can redistribute it and/or modify it +# under the terms of the GNU Affero General Public License +# as published by the Free Software Foundation, either version 3 of the +# License, or (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. +# See the GNU Affero General Public License for more details. +# +# This program is available from Source Forge: at GeneNetwork Project +# (sourceforge.net/projects/genenetwork/). +# +# Contact Drs. Robert W. Williams and Xiaodong Zhou (2010) +# at rwilliams@uthsc.edu and xzhou15@uthsc.edu +# +# +# +# This module is used by GeneNetwork project (www.genenetwork.org) +# +# Created by GeneNetwork Core Team 2010/08/10 +# +# Last updated by GeneNetwork Core Team 2010/10/20 + +import string +import time +import re +import math +from math import * + +from htmlgen import HTMLgen2 as HT + +from base import webqtlConfig + + + + +# NL, 07/27/2010. moved from webqtlForm.py +#Dict of Parents and F1 information, In the order of [F1, Mat, Pat] +ParInfo ={ +'BXH':['BHF1', 'HBF1', 'C57BL/6J', 'C3H/HeJ'], +'AKXD':['AKF1', 'KAF1', 'AKR/J', 'DBA/2J'], +'BXD':['B6D2F1', 'D2B6F1', 'C57BL/6J', 'DBA/2J'], +'BXD300':['B6D2F1', 'D2B6F1', 'C57BL/6J', 'DBA/2J'], +'B6BTBRF2':['B6BTBRF1', 'BTBRB6F1', 'C57BL/6J', 'BTBRT<+>tf/J'], +'BHHBF2':['B6HF2','HB6F2','C57BL/6J','C3H/HeJ'], +'BHF2':['B6HF2','HB6F2','C57BL/6J','C3H/HeJ'], +'B6D2F2':['B6D2F1', 'D2B6F1', 'C57BL/6J', 'DBA/2J'], +'BDF2-1999':['B6D2F2', 'D2B6F2', 'C57BL/6J', 'DBA/2J'], +'BDF2-2005':['B6D2F1', 'D2B6F1', 'C57BL/6J', 'DBA/2J'], +'CTB6F2':['CTB6F2','B6CTF2','C57BL/6J','Castaneous'], +'CXB':['CBF1', 'BCF1', 'C57BL/6ByJ', 'BALB/cByJ'], +'AXBXA':['ABF1', 'BAF1', 'C57BL/6J', 'A/J'], +'AXB':['ABF1', 'BAF1', 'C57BL/6J', 'A/J'], +'BXA':['BAF1', 'ABF1', 'C57BL/6J', 'A/J'], +'LXS':['LSF1', 'SLF1', 'ISS', 'ILS'], +'HXBBXH':['HSRBNF1', 'BNHSRF1', 'BN', 'HSR'], +'BayXSha':['BayXShaF1', 'ShaXBayF1', 'Bay-0','Shahdara'], +'ColXBur':['ColXBurF1', 'BurXColF1', 'Col-0','Bur-0'], +'ColXCvi':['ColXCviF1', 'CviXColF1', 'Col-0','Cvi'], +'SXM':['SMF1', 'MSF1', 'Steptoe','Morex'] +} + + +# NL, 07/27/2010. moved from template.py +IMGSTEP1 = HT.Image('/images/step1.gif', alt='STEP 1',border=0) #XZ, Only be used in inputPage.py +IMGSTEP2 = HT.Image('/images/step2.gif', alt='STEP 2',border=0) #XZ, Only be used in inputPage.py +IMGSTEP3 = HT.Image('/images/step3.gif', alt='STEP 3',border=0) #XZ, Only be used in inputPage.py +IMGNEXT = HT.Image('/images/arrowdown.gif', alt='NEXT',border=0) #XZ, Only be used in inputPage.py + +IMGASC = HT.Image("/images/sortup.gif", border=0) +IMGASCON = HT.Image("/images/sortupon.gif", border=0) +IMGDESC = HT.Image("/images/sortdown.gif", border=0) +IMGDESCON = HT.Image("/images/sortdownon.gif", border=0) + +""" +IMGASC = HT.Image("/images/sortup_icon.gif", border=0) +IMGASCON = HT.Image("/images/sortupon.gif", border=0) +IMGDESC = HT.Image("/images/sortdown_icon.gif", border=0) +IMGDESCON = HT.Image("/images/sortdownon.gif", border=0) +IMG_UNSORTED = HT.Image("/images/unsorted_icon.gif", border=0) +""" + +PROGRESSBAR = HT.Image('/images/waitAnima2.gif', alt='checkblue',align="middle",border=0) + +######################################### +# Accessory Functions +######################################### + +def decodeEscape(str): + a = str + pattern = re.compile('(%[0-9A-Fa-f][0-9A-Fa-f])') + match = pattern.findall(a) + matched = [] + for item in match: + if item not in matched: + a = a.replace(item, '%c' % eval("0x"+item[-2:])) + matched.append(item) + return a + +def exportData(hddn, tdata, NP = None): + for key in tdata.keys(): + _val, _var, _N = tdata[key].val, tdata[key].var, tdata[key].N + if _val != None: + hddn[key] = _val + if _var != None: + hddn['V'+key] = _var + if NP and _N != None: + hddn['N'+key] = _N + +def genShortStrainName(RISet='', input_strainName=''): + #aliasStrainDict = {'C57BL/6J':'B6','DBA/2J':'D2'} + strainName = input_strainName + if RISet != 'AXBXA': + if RISet == 'BXD300': + this_RISet = 'BXD' + elif RISet == 'BDF2-2005': + this_RISet = 'CASE05_' + else: + this_RISet = RISet + strainName = string.replace(strainName,this_RISet,'') + strainName = string.replace(strainName,'CASE','') + try: + strainName = "%02d" % int(strainName) + except: + pass + else: + strainName = string.replace(strainName,'AXB','A') + strainName = string.replace(strainName,'BXA','B') + try: + strainName = strainName[0] + "%02d" % int(strainName[1:]) + except: + pass + return strainName + +def toInt(in_str): + "Converts an arbitrary string to an unsigned integer" + start = -1 + end = -1 + for i, char in enumerate(in_str): + if char >= '0' and char <= '9': + if start < 0: + start = i + end = i+1 + else: + if start >= 0: + break + if start < end: + return int(in_str[start:end]) + else: + return -1 + +def transpose(m): + 'transpose a matrix' + n = len(m) + return [[m[j][i] for i in range(len(m[0])) for j in range(n)][k*n:k*n+n] for k in range(len(m[0]))] + +def asymTranspose(m): + 'transpose a matrix' + t = max(map(len, m)) + n = len(m) + m2 = [["-"]]*n + for i in range(n): + m2[i] = m[i] + [""]*(t- len(m[i])) + return [[m2[j][i] for i in range(len(m2[0])) for j in range(n)][k*n:k*n+n] for k in range(len(m2[0]))] + +def genRandStr(prefix = "", length=8, chars=string.letters+string.digits): + from random import choice + _str = prefix[:] + for i in range(length): + _str += choice(chars) + return _str + +def generate_session(): + import sha + return sha.new(str(time.time())).hexdigest() + +def cvt2Dict(x): + tmp = {} + for key in x.keys(): + tmp[key] = x[key] + return tmp + +def dump_session(session_obj, filename): + "It seems mod python can only cPickle most basic data type" + import cPickle + session_file = open(filename, 'wb') + #try: + # pass + #except: + # pass + cPickle.dump(session_obj, session_file) + session_file.close() + +def StringAsFloat(str): + 'Converts string to float but catches any exception and returns None' + try: + return float(str) + except: + return None + +def IntAsFloat(str): + 'Converts string to Int but catches any exception and returns None' + try: + return int(str) + except: + return None + +def FloatAsFloat(flt): + 'Converts float to string but catches any exception and returns None' + try: + return float("%2.3f" % flt) + except: + return None + +def RemoveZero(flt): + 'Converts string to float but catches any exception and returns None' + try: + if abs(flt) < 1e-6: + return None + else: + return flt + except: + return None + + +def SciFloat(d): + 'Converts string to float but catches any exception and returns None' + + try: + if abs(d) <= 1.0e-4: + return "%1.2e" % d + else: + return "%1.5f" % d + except: + return None + +###To be removed +def FloatList2String(lst): + 'Converts float list to string but catches any exception and returns None' + tt='' + try: + for item in lst: + if item == None: + tt += 'X ' + else: + tt += '%f ' % item + return tt + except: + return "" + +def ListNotNull(lst): + '''Obsolete - Use built in function any (or all or whatever) + + + Determine if the elements in a list are all null + + ''' + for item in lst: + if item is not None: + return 1 + return None + +###To be removed +def FileDataProcess(str): + 'Remove the description text from the input file if theres any' + i=0 + while i<len(str): + if str[i]<'\x7f' and str[i]>'\x20': + break + else: + i+=1 + str=str[i:] + str=string.join(string.split(str,'\000'),'') + i=string.find(str,"*****") + if i>-1: + return str[i+5:] + else: + return str + +def rank(a,lst,offset=0): + """Calculate the integer rank of a number in an array, can be used to calculate p-value""" + n = len(lst) + if n == 2: + if a <lst[0]: + return offset + elif a > lst[1]: + return offset + 2 + else: + return offset +1 + elif n == 1: + if a <lst[0]: + return offset + else: + return offset +1 + elif n== 0: + return offset + else: + mid = n/2 + if a < lst[mid]: + return rank(a,lst[:mid-1],offset) + else: + return rank(a,lst[mid:],offset+mid) + +def cmpScanResult(A,B): + try: + if A.LRS > B.LRS: + return 1 + elif A.LRS == B.LRS: + return 0 + else: + return -1 + except: + return 0 + + +def cmpScanResult2(A,B): + try: + if A.LRS < B.LRS: + return 1 + elif A.LRS == B.LRS: + return 0 + else: + return -1 + except: + return 0 + +def cmpOrder(A,B): + try: + if A[1] < B[1]: + return -1 + elif A[1] == B[1]: + return 0 + else: + return 1 + except: + return 0 + +def cmpOrder2(A,B): + try: + if A[-1] < B[-1]: + return -1 + elif A[-1] == B[-1]: + return 0 + else: + return 1 + except: + return 0 + + + + +def calRank(xVals, yVals, N): ### Zach Sloan, February 4 2010 + """ + Returns a ranked set of X and Y values. These are used when generating + a Spearman scatterplot. Bear in mind that this sets values equal to each + other as the same rank. + """ + XX = [] + YY = [] + X = [0]*len(xVals) + Y = [0]*len(yVals) + j = 0 + + for i in range(len(xVals)): + + if xVals[i] != None and yVals[i] != None: + XX.append((j, xVals[i])) + YY.append((j, yVals[i])) + j = j + 1 + + NN = len(XX) + + XX.sort(cmpOrder2) + YY.sort(cmpOrder2) + + j = 1 + rank = 0.0 + + while j < NN: + + if XX[j][1] != XX[j-1][1]: + X[XX[j-1][0]] = j + j = j+1 + + else: + jt = j+1 + ji = j + for jt in range(j+1, NN): + if (XX[jt][1] != XX[j-1][1]): + break + rank = 0.5*(j+jt) + for ji in range(j-1, jt): + X[XX[ji][0]] = rank + if (jt == NN-1): + if (XX[jt][1] == XX[j-1][1]): + X[XX[NN-1][0]] = rank + j = jt+1 + + if j == NN: + if X[XX[NN-1][0]] == 0: + X[XX[NN-1][0]] = NN + + j = 1 + rank = 0.0 + + while j < NN: + + if YY[j][1] != YY[j-1][1]: + Y[YY[j-1][0]] = j + j = j+1 + else: + jt = j+1 + ji = j + for jt in range(j+1, NN): + if (YY[jt][1] != YY[j-1][1]): + break + rank = 0.5*(j+jt) + for ji in range(j-1, jt): + Y[YY[ji][0]] = rank + if (jt == NN-1): + if (YY[jt][1] == YY[j-1][1]): + Y[YY[NN-1][0]] = rank + j = jt+1 + + if j == NN: + if Y[YY[NN-1][0]] == 0: + Y[YY[NN-1][0]] = NN + + return (X,Y) + +def calCorrelationRank(xVals,yVals,N): + """ + Calculated Spearman Ranked Correlation. The algorithm works + by setting all tied ranks to the average of those ranks (for + example, if ranks 5-10 all have the same value, each will be set + to rank 7.5). + """ + + XX = [] + YY = [] + j = 0 + + for i in range(len(xVals)): + if xVals[i]!= None and yVals[i]!= None: + XX.append((j,xVals[i])) + YY.append((j,yVals[i])) + j = j+1 + + NN = len(XX) + if NN <6: + return (0.0,NN) + XX.sort(cmpOrder2) + YY.sort(cmpOrder2) + X = [0]*NN + Y = [0]*NN + + j = 1 + rank = 0.0 + t = 0.0 + sx = 0.0 + + while j < NN: + + if XX[j][1] != XX[j-1][1]: + X[XX[j-1][0]] = j + j = j+1 + + else: + jt = j+1 + ji = j + for jt in range(j+1, NN): + if (XX[jt][1] != XX[j-1][1]): + break + rank = 0.5*(j+jt) + for ji in range(j-1, jt): + X[XX[ji][0]] = rank + t = jt-j + sx = sx + (t*t*t-t) + if (jt == NN-1): + if (XX[jt][1] == XX[j-1][1]): + X[XX[NN-1][0]] = rank + j = jt+1 + + if j == NN: + if X[XX[NN-1][0]] == 0: + X[XX[NN-1][0]] = NN + + j = 1 + rank = 0.0 + t = 0.0 + sy = 0.0 + + while j < NN: + + if YY[j][1] != YY[j-1][1]: + Y[YY[j-1][0]] = j + j = j+1 + else: + jt = j+1 + ji = j + for jt in range(j+1, NN): + if (YY[jt][1] != YY[j-1][1]): + break + rank = 0.5*(j+jt) + for ji in range(j-1, jt): + Y[YY[ji][0]] = rank + t = jt - j + sy = sy + (t*t*t-t) + if (jt == NN-1): + if (YY[jt][1] == YY[j-1][1]): + Y[YY[NN-1][0]] = rank + j = jt+1 + + if j == NN: + if Y[YY[NN-1][0]] == 0: + Y[YY[NN-1][0]] = NN + + D = 0.0 + + for i in range(NN): + D += (X[i]-Y[i])*(X[i]-Y[i]) + + fac = (1.0 -sx/(NN*NN*NN-NN))*(1.0-sy/(NN*NN*NN-NN)) + + return ((1-(6.0/(NN*NN*NN-NN))*(D+(sx+sy)/12.0))/math.sqrt(fac),NN) + + +def calCorrelationRankText(dbdata,userdata,N): ### dcrowell = David Crowell, July 2008 + """Calculates correlation ranks with data formatted from the text file. + dbdata, userdata are lists of strings. N is an int. Returns a float. + Used by correlationPage""" + XX = [] + YY = [] + j = 0 + for i in range(N): + if (dbdata[i]!= None and userdata[i]!=None) and (dbdata[i]!= 'None' and userdata[i]!='None'): + XX.append((j,float(dbdata[i]))) + YY.append((j,float(userdata[i]))) + j += 1 + NN = len(XX) + if NN <6: + return (0.0,NN) + XX.sort(cmpOrder2) + YY.sort(cmpOrder2) + X = [0]*NN + Y = [0]*NN + + j = 1 + rank = 0.0 + t = 0.0 + sx = 0.0 + + while j < NN: + + if XX[j][1] != XX[j-1][1]: + X[XX[j-1][0]] = j + j = j+1 + + else: + jt = j+1 + ji = j + for jt in range(j+1, NN): + if (XX[jt][1] != XX[j-1][1]): + break + rank = 0.5*(j+jt) + for ji in range(j-1, jt): + X[XX[ji][0]] = rank + t = jt-j + sx = sx + (t*t*t-t) + if (jt == NN-1): + if (XX[jt][1] == XX[j-1][1]): + X[XX[NN-1][0]] = rank + j = jt+1 + + if j == NN: + if X[XX[NN-1][0]] == 0: + X[XX[NN-1][0]] = NN + + j = 1 + rank = 0.0 + t = 0.0 + sy = 0.0 + + while j < NN: + + if YY[j][1] != YY[j-1][1]: + Y[YY[j-1][0]] = j + j = j+1 + else: + jt = j+1 + ji = j + for jt in range(j+1, NN): + if (YY[jt][1] != YY[j-1][1]): + break + rank = 0.5*(j+jt) + for ji in range(j-1, jt): + Y[YY[ji][0]] = rank + t = jt - j + sy = sy + (t*t*t-t) + if (jt == NN-1): + if (YY[jt][1] == YY[j-1][1]): + Y[YY[NN-1][0]] = rank + j = jt+1 + + if j == NN: + if Y[YY[NN-1][0]] == 0: + Y[YY[NN-1][0]] = NN + + D = 0.0 + + for i in range(NN): + D += (X[i]-Y[i])*(X[i]-Y[i]) + + fac = (1.0 -sx/(NN*NN*NN-NN))*(1.0-sy/(NN*NN*NN-NN)) + + return ((1-(6.0/(NN*NN*NN-NN))*(D+(sx+sy)/12.0))/math.sqrt(fac),NN) + + + +def calCorrelation(dbdata,userdata,N): + X = [] + Y = [] + for i in range(N): + if dbdata[i]!= None and userdata[i]!= None: + X.append(dbdata[i]) + Y.append(userdata[i]) + NN = len(X) + if NN <6: + return (0.0,NN) + sx = reduce(lambda x,y:x+y,X,0.0) + sy = reduce(lambda x,y:x+y,Y,0.0) + meanx = sx/NN + meany = sy/NN + xyd = 0.0 + sxd = 0.0 + syd = 0.0 + for i in range(NN): + xyd += (X[i] - meanx)*(Y[i]-meany) + sxd += (X[i] - meanx)*(X[i] - meanx) + syd += (Y[i] - meany)*(Y[i] - meany) + try: + corr = xyd/(sqrt(sxd)*sqrt(syd)) + except: + corr = 0 + return (corr,NN) + +def calCorrelationText(dbdata,userdata,N): ### dcrowell July 2008 + """Calculates correlation coefficients with values formatted from text files. dbdata, userdata are lists of strings. N is an int. Returns a float + Used by correlationPage""" + X = [] + Y = [] + for i in range(N): + #if (dbdata[i]!= None and userdata[i]!= None) and (dbdata[i]!= 'None' and userdata[i]!= 'None'): + # X.append(float(dbdata[i])) + # Y.append(float(userdata[i])) + if dbdata[i] == None or dbdata[i] == 'None' or userdata[i] == None or userdata[i] == 'None': + continue + else: + X.append(float(dbdata[i])) + Y.append(float(userdata[i])) + NN = len(X) + if NN <6: + return (0.0,NN) + sx = sum(X) + sy = sum(Y) + meanx = sx/float(NN) + meany = sy/float(NN) + xyd = 0.0 + sxd = 0.0 + syd = 0.0 + for i in range(NN): + x1 = X[i]-meanx + y1 = Y[i]-meany + xyd += x1*y1 + sxd += x1**2 + syd += y1**2 + try: + corr = xyd/(sqrt(sxd)*sqrt(syd)) + except: + corr = 0 + return (corr,NN) + + +def readLineCSV(line): ### dcrowell July 2008 + """Parses a CSV string of text and returns a list containing each element as a string. + Used by correlationPage""" + returnList = line.split('","') + returnList[-1]=returnList[-1][:-2] + returnList[0]=returnList[0][1:] + return returnList + + +def cmpCorr(A,B): + try: + if abs(A[1]) < abs(B[1]): + return 1 + elif abs(A[1]) == abs(B[1]): + return 0 + else: + return -1 + except: + return 0 + +def cmpLitCorr(A,B): + try: + if abs(A[3]) < abs(B[3]): return 1 + elif abs(A[3]) == abs(B[3]): + if abs(A[1]) < abs(B[1]): return 1 + elif abs(A[1]) == abs(B[1]): return 0 + else: return -1 + else: return -1 + except: + return 0 + +def cmpPValue(A,B): + try: + if A.corrPValue < B.corrPValue: + return -1 + elif A.corrPValue == B.corrPValue: + if abs(A.corr) > abs(B.corr): + return -1 + elif abs(A.corr) < abs(B.corr): + return 1 + else: + return 0 + else: + return 1 + except: + return 0 + +def cmpEigenValue(A,B): + try: + if A[0] > B[0]: + return -1 + elif A[0] == B[0]: + return 0 + else: + return 1 + except: + return 0 + + +def cmpLRSFull(A,B): + try: + if A[0] < B[0]: + return -1 + elif A[0] == B[0]: + return 0 + else: + return 1 + except: + return 0 + +def cmpLRSInteract(A,B): + try: + if A[1] < B[1]: + return -1 + elif A[1] == B[1]: + return 0 + else: + return 1 + except: + return 0 + + +def cmpPos(A,B): + try: + try: + AChr = int(A.chr) + except: + AChr = 20 + try: + BChr = int(B.chr) + except: + BChr = 20 + if AChr > BChr: + return 1 + elif AChr == BChr: + if A.mb > B.mb: + return 1 + if A.mb == B.mb: + return 0 + else: + return -1 + else: + return -1 + except: + return 0 + +def cmpGenoPos(A,B): + try: + A1 = A.chr + B1 = B.chr + try: + A1 = int(A1) + except: + A1 = 25 + try: + B1 = int(B1) + except: + B1 = 25 + if A1 > B1: + return 1 + elif A1 == B1: + if A.mb > B.mb: + return 1 + if A.mb == B.mb: + return 0 + else: + return -1 + else: + return -1 + except: + return 0 + +#XZhou: Must use "BINARY" to enable case sensitive comparison. +def authUser(name,password,db, encrypt=None): + try: + if encrypt: + query = 'SELECT privilege, id,name,password, grpName FROM User WHERE name= BINARY \'%s\' and password= BINARY \'%s\'' % (name,password) + else: + query = 'SELECT privilege, id,name,password, grpName FROM User WHERE name= BINARY \'%s\' and password= BINARY SHA(\'%s\')' % (name,password) + db.execute(query) + records = db.fetchone() + if not records: + raise ValueError + return records#(privilege,id,name,password,grpName) + except: + return (None, None, None, None, None) + + +def hasAccessToConfidentialPhenotypeTrait(privilege, userName, authorized_users): + access_to_confidential_phenotype_trait = 0 + if webqtlConfig.USERDICT[privilege] > webqtlConfig.USERDICT['user']: + access_to_confidential_phenotype_trait = 1 + else: + AuthorizedUsersList=map(string.strip, string.split(authorized_users, ',')) + if AuthorizedUsersList.__contains__(userName): + access_to_confidential_phenotype_trait = 1 + return access_to_confidential_phenotype_trait + + +class VisualizeException(Exception): + def __init__(self, message): + self.message = message + def __str__(self): + return self.message + +# safeConvert : (string -> A) -> A -> A +# to convert a string to type A, using the supplied default value +# if the given conversion function doesn't work +def safeConvert(f, value, default): + try: + return f(value) + except: + return default + +# safeFloat : string -> float -> float +# to convert a string to a float safely +def safeFloat(value, default): + return safeConvert(float, value, default) + +# safeInt: string -> int -> int +# to convert a string to an int safely +def safeInt(value, default): + return safeConvert(int, value, default) + +# safeString : string -> (arrayof string) -> string -> string +# if a string is not in a list of strings to pick a default value +# for that string +def safeString(value, validChoices, default): + if value in validChoices: + return value + else: + return default + +# yesNoToInt: string -> int +# map "yes" -> 1 and "no" -> 0 +def yesNoToInt(value): + if value == "yes": + return 1 + elif value == "no": + return 0 + else: + return None + +# IntToYesNo: int -> string +# map 1 -> "yes" and 0 -> "no" +def intToYesNo(value): + if value == 1: + return "yes" + elif value == 0: + return "no" + else: + return None + +def formatField(name): + name = name.replace("_", " ") + name = name.title() + #name = name.replace("Mb Mm6", "Mb"); + return name.replace("Id", "ID") + +#XZ, 03/27/2009: This function is very specific. +#It is used by AJAX_table.py, correlationPage.py and dataPage.py + + +def genTableObj(tblobj=None, file="", sortby = ("", ""), tableID = "sortable", addIndex = "1", hiddenColumns=[]): + header = tblobj['header'] + body = tblobj['body'] + field, order = sortby + + #ZAS 9/12/2011 - The hiddenColumns array needs to be converted into a string so they can be placed into the javascript of each up/down button + hiddenColumnsString = ",".join(hiddenColumns) + + tbl = HT.TableLite(Class="collap b2", cellspacing=1, cellpadding=5) + + hiddenColumnIdx = [] #indices of columns to hide + idx = -1 + last_idx = 0 #ZS: This is the index of the last item in the regular table header (without any extra parameters). It is used to determine the index of each extra parameter. + for row in header: + hr = HT.TR() + for i, item in enumerate(row): + if (item.text == '') or (item.text not in hiddenColumns): + if item.sort and item.text: + down = HT.Href("javascript:xmlhttpPost('%smain.py?FormID=AJAX_table', '%s', 'sort=%s&order=down&file=%s&tableID=%s&addIndex=%s&hiddenColumns=%s')" % (webqtlConfig.CGIDIR, tableID, item.text, file, tableID, addIndex, hiddenColumnsString),IMGDESC) + up = HT.Href("javascript:xmlhttpPost('%smain.py?FormID=AJAX_table', '%s', 'sort=%s&order=up&file=%s&tableID=%s&addIndex=%s&hiddenColumns=%s')" % (webqtlConfig.CGIDIR, tableID, item.text, file, tableID, addIndex, hiddenColumnsString),IMGASC) + if item.text == field: + idx = item.idx + last_idx = idx + if order == 'up': + up = IMGASCON + elif order == 'down': + down = IMGDESCON + item.html.append(HT.Div(up, down, style="float: bottom;")) + hr.append(item.html) + else: + hiddenColumnIdx.append(i) + tbl.append(hr) + + for i, row in enumerate(body): + for j, item in enumerate(row): + if order == 'down': + if (item.val == '' or item.val == 'x' or item.val == 'None'): + item.val = 0 + if order == 'up': + if (item.val == '' or item.val == 'x' or item.val == 'None'): + item.val = 'zzzzz' + + if idx >= 0: + if order == 'down': + body.sort(lambda A, B: cmp(B[idx].val, A[idx].val), key=natsort_key) + elif order == 'up': + body.sort(lambda A, B: cmp(A[idx].val, B[idx].val), key=natsort_key) + else: + pass + + for i, row in enumerate(body): + hr = HT.TR(Id = row[0].text) + for j, item in enumerate(row): + if (j not in hiddenColumnIdx): + if j == 0: + if addIndex == "1": + item.html.contents = [i+1] + item.html.contents + hr.append(item.html) + tbl.append(hr) + + return tbl + +def natsort_key(string): + r = [] + for c in string: + try: + c = int(c) + try: r[-1] = r[-1] * 10 + c + except: r.append(c) + except: + r.append(c) + return r |