forked from tomaszswiderski/Spreadsheet-Table-for-ReportLab
-
Notifications
You must be signed in to change notification settings - Fork 0
/
spreadsheettable.py
1392 lines (1261 loc) · 58.9 KB
/
spreadsheettable.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# -*- coding: utf-8 -*-
__author__ = u'Tomasz Świderski <contact@tomaszswiderski.com>'
__copyright__ = u'Copyright (c) 2010 Tomasz Świderski'
from reportlab.platypus.tables import *
from reportlab.platypus.tables import (_rowLen, _calc_pc, _hLine, _multiLine,
_convert2int, _endswith, _isLineCommand, _setCellStyle)
from formula import Formula
def spanFixDim(V0,V,spanCons,FUZZ=rl_config._FUZZ):
#assign required space to variable rows equally to existing calculated values
M = {}
for (x0,x1),v in spanCons.iteritems():
t = sum([V[x]+M.get(x,0) for x in xrange(x0,x1+1)])
if t>=v-FUZZ: continue #already good enough
X = [x for x in xrange(x0,x1+1) if V0[x] is None] #variable candidates
if not X: continue #something wrong here mate
v -= t
v /= float(len(X))
for x in X:
M[x] = M.get(x,0) + v
for x,v in M.iteritems():
V[x] += v
class SpreadsheetTable(Flowable):
def __init__(self, data, colWidths=None, rowHeights=None, style=None,
repeatRows=0, repeatCols=0, splitByRow=1, emptyTableAction=None,
ident=None, hAlign='CENTER', vAlign='MIDDLE', normalizedData=0,
cellStyles=None, activeRows=None, repeatRowsB=0):
self.ident = ident
self.hAlign = hAlign
self.vAlign = vAlign
if not isinstance(data,(tuple,list)):
raise ValueError("%s invalid data type" % self.identity())
self._nrows = nrows = len(data)
self.repeatRows = repeatRows
self._repeatRowsB = repeatRowsB
if activeRows is None:
activeRows = (None, nrows - repeatRowsB)
self._activeRows = activeRows
self._cellvalues = []
_seqCW = isinstance(colWidths,(tuple,list))
_seqRH = isinstance(rowHeights,(tuple,list))
if nrows: self._ncols = ncols = max(map(_rowLen,data))
elif colWidths and _seqCW: ncols = len(colWidths)
else: ncols = 0
if not emptyTableAction: emptyTableAction = rl_config.emptyTableAction
if not (nrows and ncols):
if emptyTableAction=='error':
raise ValueError("%s must have at least a row and column" % self.identity())
elif emptyTableAction=='indicate':
self.__class__ = Preformatted
global _emptyTableStyle
if '_emptyTableStyle' not in globals().keys():
_emptyTableStyle = ParagraphStyle('_emptyTableStyle')
_emptyTableStyle.textColor = colors.red
_emptyTableStyle.backColor = colors.yellow
Preformatted.__init__(self,'%s(%d,%d)' % (self.__class__.__name__,nrows,ncols), _emptyTableStyle)
elif emptyTableAction=='ignore':
self.__class__ = Spacer
Spacer.__init__(self,0,0)
else:
raise ValueError('%s bad emptyTableAction: "%s"' % (self.identity(),emptyTableAction))
return
# we need a cleanup pass to ensure data is strings - non-unicode and non-null
if normalizedData:
self._cellvalues = data
else:
self._cellvalues = data = self.normalizeData(data)
if not _seqCW: colWidths = ncols*[colWidths]
elif len(colWidths)!=ncols:
if rl_config.allowShortTableRows and isinstance(colWidths,list):
n = len(colWidths)
if n<ncols:
colWidths[n:] = (ncols-n)*[colWidths[-1]]
else:
colWidths = colWidths[:ncols]
else:
raise ValueError("%s data error - %d columns in data but %d in column widths" % (self.identity(),ncols, len(colWidths)))
if not _seqRH: rowHeights = nrows*[rowHeights]
elif len(rowHeights) != nrows:
raise ValueError("%s data error - %d rows in data but %d in row heights" % (self.identity(),nrows, len(rowHeights)))
for i,d in enumerate(data):
n = len(d)
if n!=ncols:
if rl_config.allowShortTableRows and isinstance(d,list):
d[n:] = (ncols-n)*['']
else:
raise ValueError("%s expected %d not %d columns in row %d!" % (self.identity(),ncols,n,i))
self._argH = self._rowHeights = rowHeights
self._colWidths = self._argW = colWidths
if cellStyles is None:
cellrows = []
for i in xrange(nrows):
cellcols = []
for j in xrange(ncols):
cellcols.append(CellStyle(`(i,j)`))
cellrows.append(cellcols)
self._cellStyles = cellrows
else:
self._cellStyles = cellStyles
self._bkgrndcmds = []
self._linecmds = []
self._spanCmds = []
self._nosplitCmds = []
self.repeatCols = repeatCols
self.splitByRow = splitByRow
if style:
self.setStyle(style)
def __repr__(self):
"incomplete, but better than nothing"
r = getattr(self,'_rowHeights','[unknown]')
c = getattr(self,'_colWidths','[unknown]')
cv = getattr(self,'_cellvalues','[unknown]')
import pprint
cv = pprint.pformat(cv)
cv = cv.replace("\n", "\n ")
return "%s(\n rowHeights=%s,\n colWidths=%s,\n%s\n) # end table" % (self.__class__.__name__,r,c,cv)
def normalizeData(self, data):
"""Takes a block of input data (list of lists etc.) and
- coerces unicode strings to non-unicode UTF8
- coerces nulls to ''
"""
def normCell(stuff):
if stuff is None:
return ''
elif isinstance(stuff,unicode):
return stuff.encode('utf8')
else:
return stuff
outData = []
for row in data:
outRow = [normCell(cell) for cell in row]
outData.append(outRow)
return outData
def identity(self, maxLen=30):
'''Identify our selves as well as possible'''
if self.ident: return self.ident
vx = None
nr = getattr(self,'_nrows','unknown')
nc = getattr(self,'_ncols','unknown')
cv = getattr(self,'_cellvalues',None)
if cv and 'unknown' not in (nr,nc):
b = 0
for i in xrange(nr):
for j in xrange(nc):
v = cv[i][j]
if isinstance(v,(list,tuple,Flowable)):
if not isinstance(v,(tuple,list)): v = (v,)
r = ''
for vij in v:
r = vij.identity(maxLen)
if r and r[-4:]!='>...':
break
if r and r[-4:]!='>...':
ix, jx, vx, b = i, j, r, 1
else:
v = v is None and '' or str(v)
ix, jx, vx = i, j, v
b = (vx and isinstance(v,basestring)) and 1 or 0
if maxLen: vx = vx[:maxLen]
if b: break
if b: break
if vx:
vx = ' with cell(%d,%d) containing\n%s' % (ix,jx,repr(vx))
else:
vx = '...'
return "<%s@0x%8.8X %s rows x %s cols>%s" % (self.__class__.__name__, id(self), nr, nc, vx)
def _listCellGeom(self, V,w,s,W=None,H=None,aH=72000):
if not V: return 0,0
aW = w - s.leftPadding - s.rightPadding
aH = aH - s.topPadding - s.bottomPadding
t = 0
w = 0
canv = getattr(self,'canv',None)
sb0 = None
for v in V:
vw, vh = v.wrapOn(canv, aW, aH)
sb = v.getSpaceBefore()
sa = v.getSpaceAfter()
if W is not None: W.append(vw)
if H is not None: H.append(vh)
w = max(w,vw)
t += vh + sa + sb
if sb0 is None:
sb0 = sb
return w, t - sb0 - sa
def _listValueWidth(self,V,aH=72000,aW=72000):
if not V: return 0,0
t = 0
w = 0
canv = getattr(self,'canv',None)
return max([v.wrapOn(canv,aW,aH)[0] for v in V])
def _calc_width(self,availWidth,W=None):
if getattr(self,'_width_calculated_once',None): return
#comments added by Andy to Robin's slightly terse variable names
if not W: W = _calc_pc(self._argW,availWidth) #widths array
if None in W: #some column widths are not given
canv = getattr(self,'canv',None)
saved = None
if self._spanCmds:
colSpanCells = self._colSpanCells
spanRanges = self._spanRanges
else:
colSpanCells = ()
spanRanges = {}
spanCons = {}
if W is self._argW:
W0 = W
W = W[:]
else:
W0 = W[:]
V = self._cellvalues
S = self._cellStyles
while None in W:
j = W.index(None) #find first unspecified column
w = 0
for i,Vi in enumerate(V):
v = Vi[j]
s = S[i][j]
ji = j,i
span = spanRanges.get(ji,None)
if ji in colSpanCells and not span: #if the current cell is part of a spanned region,
t = 0.0 #assume a zero size.
else:#work out size
t = self._elementWidth(v, s, (j, i))
if t is None:
raise ValueError("Flowable %s in cell(%d,%d) can't have auto width\n%s" % (v.identity(30),i,j,self.identity(30)))
t += s.leftPadding+s.rightPadding
if span:
c0 = span[0]
c1 = span[2]
if c0!=c1:
x = c0,c1
spanCons[x] = max(spanCons.get(x,t),t)
t = 0
if t>w: w = t #record a new maximum
W[j] = w
if spanCons:
spanFixDim(W0,W,spanCons)
self._colWidths = W
width = 0
self._colpositions = [0] #index -1 is right side boundary; we skip when processing cells
for w in W:
width = width + w
self._colpositions.append(width)
self._width = width
self._width_calculated_once = 1
def _elementWidth(self, v, s, cellcoord):
if isinstance(v,(list,tuple)):
w = 0
for e in v:
ew = self._elementWidth(e,s)
if ew is None: return None
w = max(w,ew)
return w
elif isinstance(v,Flowable) and v._fixedWidth:
if hasattr(v, 'width') and isinstance(v.width,(int,float)): return v.width
if hasattr(v, 'drawWidth') and isinstance(v.drawWidth,(int,float)): return v.drawWidth
elif isinstance(v, Formula):
v = v.get_max_value(self._cellvalues, self.repeatRows,
self._repeatRowsB, cellcoord)
# Even if something is fixedWidth, the attribute to check is not
# necessarily consistent (cf. Image.drawWidth). Therefore, we'll
# be extra-careful and fall through to this code if necessary.
if hasattr(v, 'minWidth'):
try:
w = v.minWidth() # should be all flowables
if isinstance(w,(float,int)): return w
except AttributeError:
pass
v = (v is not None and str(v) or '').split("\n")
fontName = s.fontname
fontSize = s.fontsize
return max([stringWidth(x,fontName,fontSize) for x in v])
def _calc_height(self, availHeight, availWidth, H=None):
H0 = self._argH
H = self._rowHeights
W = self._colWidths
hmax = lim = len(H)
if None in H:
canv = getattr(self,'canv',None)
saved = None
#get a handy list of any cells which span rows. should be ignored for sizing
if self._spanCmds:
rowSpanCells = self._rowSpanCells
colSpanCells = self._colSpanCells
spanRanges = self._spanRanges
colpositions = self._colpositions
else:
rowSpanCells = colSpanCells = ()
spanRanges = {}
if canv: saved = canv._fontname, canv._fontsize, canv._leading
spanCons = {}
FUZZ = rl_config._FUZZ
while None in H:
i = H.index(None)
V = self._cellvalues[i] # values for row i
S = self._cellStyles[i] # styles for row i
h = 0
j = 0
for j,(v, s, w) in enumerate(zip(V, S, W)): # value, style, width (lengths must match)
ji = j,i
span = spanRanges.get(ji,None)
if ji in rowSpanCells and not span:
continue # don't count it, it's either occluded or unreliable
if isinstance(v,(tuple,list,Flowable)):
if isinstance(v,Flowable): v = (v,)
if w is None and not self._canGetWidth(v):
raise ValueError("Flowable %s in cell(%d,%d) can't have auto width in\n%s" % (v[0].identity(30),i,j,self.identity(30)))
if canv: canv._fontname, canv._fontsize, canv._leading = s.fontname, s.fontsize, s.leading or 1.2*s.fontsize
if ji in colSpanCells:
if not span: continue
w = max(colpositions[span[2]+1]-colpositions[span[0]],w)
dW,t = self._listCellGeom(v,w or self._listValueWidth(v),s)
if canv: canv._fontname, canv._fontsize, canv._leading = saved
dW = dW + s.leftPadding + s.rightPadding
if not rl_config.allowTableBoundsErrors and dW>w:
from reportlab.platypus.doctemplate import LayoutError
raise LayoutError("Flowable %s (%sx%s points) too wide for cell(%d,%d) (%sx* points) in\n%s" % (v[0].identity(30),fp_str(dW),fp_str(t),i,j, fp_str(w), self.identity(30)))
else:
v = (v is not None and str(v) or '').split("\n")
t = (s.leading or 1.2*s.fontSize)*len(v)
t += s.bottomPadding+s.topPadding
if span:
r0 = span[1]
r1 = span[3]
if r0!=r1:
x = r0,r1
spanCons[x] = max(spanCons.get(x,t),t)
t = 0
if t>h: h = t #record a new maximum
H[i] = h
if spanCons:
spanFixDim(H0,H,spanCons)
hmax = self._activeRows[1]
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
height = self._height = sum(H[:self.repeatRows] +
H[activeRows0:hmax] +
H[self._nrows-self._repeatRowsB:])
self._rowpositions = [height] # index 0 is actually topline; we skip when processing cells
for h in H[:self.repeatRows] + H[activeRows0:hmax] + H[self._nrows-self._repeatRowsB:]:
height = height - h
self._rowpositions.append(height)
assert abs(height)<1e-8, 'Internal height error'
def _calc(self, availWidth, availHeight):
#if hasattr(self,'_width'): return
#in some cases there are unsizable things in
#cells. If so, apply a different algorithm
#and assign some withs in a less (thanks to Gary Poster) dumb way.
#this CHANGES the widths array.
if (None in self._colWidths or '*' in self._colWidths) and self._hasVariWidthElements():
W = self._calcPreliminaryWidths(availWidth) #widths
else:
W = None
# need to know which cells are part of spanned
# ranges, so _calc_height and _calc_width can ignore them
# in sizing
if self._spanCmds:
self._calcSpanRanges()
if self._nosplitCmds:
self._calcNoSplitRanges()
# calculate the full table width
self._calc_width(availWidth,W=W)
# calculate the full table height
self._calc_height(availHeight,availWidth)
if self._spanCmds:
#now work out the actual rect for each spanned cell from the underlying grid
self._calcSpanRects()
def _hasVariWidthElements(self, upToRow=None):
"""Check for flowables in table cells and warn up front.
Allow a couple which we know are fixed size such as
images and graphics."""
if upToRow is None: upToRow = self._nrows
for row in xrange(min(self._nrows, upToRow)):
for col in xrange(self._ncols):
value = self._cellvalues[row][col]
if not self._canGetWidth(value):
return 1
return 0
def _canGetWidth(self, thing):
"Can we work out the width quickly?"
if isinstance(thing,(list, tuple)):
for elem in thing:
if not self._canGetWidth(elem):
return 0
return 1
elif isinstance(thing, Flowable):
return thing._fixedWidth # must loosen this up
else: #str, number, None etc.
#anything else gets passed to str(...)
# so should be sizable
return 1
def _calcPreliminaryWidths(self, availWidth):
"""Fallback algorithm for when main one fails.
Where exact width info not given but things like
paragraphs might be present, do a preliminary scan
and assign some best-guess values."""
W = list(self._argW) # _calc_pc(self._argW,availWidth)
verbose = 0
totalDefined = 0.0
percentDefined = 0
percentTotal = 0
numberUndefined = 0
numberGreedyUndefined = 0
for w in W:
if w is None:
numberUndefined += 1
elif w == '*':
numberUndefined += 1
numberGreedyUndefined += 1
elif _endswith(w,'%'):
percentDefined += 1
percentTotal += float(w[:-1])
else:
assert isinstance(w,(int,float))
totalDefined = totalDefined + w
if verbose: print 'prelim width calculation. %d columns, %d undefined width, %0.2f units remain' % (
self._ncols, numberUndefined, availWidth - totalDefined)
#check columnwise in each None column to see if they are sizable.
given = []
sizeable = []
unsizeable = []
minimums = {}
totalMinimum = 0
elementWidth = self._elementWidth
for colNo in xrange(self._ncols):
w = W[colNo]
if w is None or w=='*' or _endswith(w,'%'):
siz = 1
current = final = None
for rowNo in xrange(self._nrows):
value = self._cellvalues[rowNo][colNo]
style = self._cellStyles[rowNo][colNo]
new = elementWidth(value,style, (colNo, rowNo))
new += style.leftPadding + style.rightPadding
final = max(current, new)
current = new
siz = siz and self._canGetWidth(value) # irrelevant now?
if siz:
sizeable.append(colNo)
else:
unsizeable.append(colNo)
minimums[colNo] = final
totalMinimum += final
else:
given.append(colNo)
if len(given) == self._ncols:
return
if verbose: print 'predefined width: ',given
if verbose: print 'uncomputable width: ',unsizeable
if verbose: print 'computable width: ',sizeable
# how much width is left:
remaining = availWidth - (totalMinimum + totalDefined)
if remaining > 0:
# we have some room left; fill it.
definedPercentage = (totalDefined/availWidth)*100
percentTotal += definedPercentage
if numberUndefined and percentTotal < 100:
undefined = numberGreedyUndefined or numberUndefined
defaultWeight = (100-percentTotal)/undefined
percentTotal = 100
defaultDesired = (defaultWeight/percentTotal)*availWidth
else:
defaultWeight = defaultDesired = 1
# we now calculate how wide each column wanted to be, and then
# proportionately shrink that down to fit the remaining available
# space. A column may not shrink less than its minimum width,
# however, which makes this a bit more complicated.
desiredWidths = []
totalDesired = 0
effectiveRemaining = remaining
for colNo, minimum in minimums.items():
w = W[colNo]
if _endswith(w,'%'):
desired = (float(w[:-1])/percentTotal)*availWidth
elif w == '*':
desired = defaultDesired
else:
desired = not numberGreedyUndefined and defaultDesired or 1
if desired <= minimum:
W[colNo] = minimum
else:
desiredWidths.append(
(desired-minimum, minimum, desired, colNo))
totalDesired += desired
effectiveRemaining += minimum
if desiredWidths: # else we're done
# let's say we have two variable columns. One wanted
# 88 points, and one wanted 264 points. The first has a
# minWidth of 66, and the second of 55. We have 71 points
# to divide up in addition to the totalMinimum (i.e.,
# remaining==71). Our algorithm tries to keep the proportion
# of these variable columns.
#
# To do this, we add up the minimum widths of the variable
# columns and the remaining width. That's 192. We add up the
# totalDesired width. That's 352. That means we'll try to
# shrink the widths by a proportion of 192/352--.545454.
# That would make the first column 48 points, and the second
# 144 points--adding up to the desired 192.
#
# Unfortunately, that's too small for the first column. It
# must be 66 points. Therefore, we go ahead and save that
# column width as 88 points. That leaves (192-88==) 104
# points remaining. The proportion to shrink the remaining
# column is (104/264), which, multiplied by the desired
# width of 264, is 104: the amount assigned to the remaining
# column.
proportion = effectiveRemaining/totalDesired
# we sort the desired widths by difference between desired and
# and minimum values, a value called "disappointment" in the
# code. This means that the columns with a bigger
# disappointment will have a better chance of getting more of
# the available space.
desiredWidths.sort()
finalSet = []
for disappointment, minimum, desired, colNo in desiredWidths:
adjusted = proportion * desired
if adjusted < minimum:
W[colNo] = minimum
totalDesired -= desired
effectiveRemaining -= minimum
if totalDesired:
proportion = effectiveRemaining/totalDesired
else:
finalSet.append((minimum, desired, colNo))
for minimum, desired, colNo in finalSet:
adjusted = proportion * desired
assert adjusted >= minimum
W[colNo] = adjusted
else:
for colNo, minimum in minimums.items():
W[colNo] = minimum
if verbose: print 'new widths are:', W
self._argW = self._colWidths = W
return W
def minWidth(self):
W = list(self._argW)
width = 0
elementWidth = self._elementWidth
rowNos = xrange(self._nrows)
values = self._cellvalues
styles = self._cellStyles
for colNo in xrange(len(W)):
w = W[colNo]
if w is None or w=='*' or _endswith(w,'%'):
final = 0
for rowNo in rowNos:
value = values[rowNo][colNo]
style = styles[rowNo][colNo]
new = (elementWidth(value, style, (colNo, rowNo)) +
style.leftPadding + style.rightPadding)
final = max(final, new)
width += final
else:
width += float(w)
return width # XXX + 1/2*(left and right border widths)
def _calcSpanRanges(self):
"""
Work out rects for tables which do row and column spanning.
This creates some mappings to let the later code determine
if a cell is part of a "spanned" range.
self._spanRanges shows the 'coords' in integers of each
'cell range', or None if it was clobbered:
(col, row) -> (col0, row0, col1, row1)
Any cell not in the key is not part of a spanned region.
This method use absolute data positions so its result can
be reused after split.
"""
# Checks if span ranges are already computed.
if getattr(self, '_spanRanges', None) is not None:
return
self._spanRanges = spanRanges = {}
for x in xrange(self._ncols):
for y in xrange(self._nrows):
spanRanges[x,y] = (x, y, x, y)
self._colSpanCells = []
self._rowSpanCells = []
csa = self._colSpanCells.append
rsa = self._rowSpanCells.append
for (cmd, start, stop) in self._spanCmds:
x0, y0 = start
x1, y1 = stop
if x0!=x1 or y0!=y1:
if x0!=x1: #column span
for y in xrange(y0, y1+1):
for x in xrange(x0,x1+1):
csa((x,y))
if y0!=y1: #row span
for y in xrange(y0, y1+1):
for x in xrange(x0,x1+1):
rsa((x,y))
for y in xrange(y0, y1+1):
for x in xrange(x0,x1+1):
spanRanges[x,y] = None
# set the main entry
spanRanges[x0,y0] = (x0, y0, x1, y1)
def _calcNoSplitRanges(self):
"""
This creates some mappings to let the later code determine
if a cell is part of a "nosplit" range.
self._nosplitRanges shows the 'coords' in integers of each
'cell range', or None if it was clobbered:
(col, row) -> (col0, row0, col1, row1)
Any cell not in the key is not part of a spanned region
"""
# Checks if nosplit ranges are already computed.
if getattr(self, '_nosplitRanges', None) is not None:
return
self._nosplitRanges = nosplitRanges = {}
for x in xrange(self._ncols):
for y in xrange(self._nrows):
nosplitRanges[x,y] = (x, y, x, y)
self._colNoSplitCells = []
self._rowNoSplitCells = []
csa = self._colNoSplitCells.append
rsa = self._rowNoSplitCells.append
for (cmd, start, stop) in self._nosplitCmds:
x0, y0 = start
x1, y1 = stop
if x0!=x1 or y0!=y1:
#column span
if x0!=x1:
for y in xrange(y0, y1+1):
for x in xrange(x0,x1+1):
csa((x,y))
#row span
if y0!=y1:
for y in xrange(y0, y1+1):
for x in xrange(x0,x1+1):
rsa((x,y))
for y in xrange(y0, y1+1):
for x in xrange(x0,x1+1):
nosplitRanges[x,y] = None
# set the main entry
nosplitRanges[x0,y0] = (x0, y0, x1, y1)
def _calcSpanRects(self):
"""
Work out rects for tables which do row and column spanning.
Based on self._spanRanges, which is already known,
and the widths which were given or previously calculated,
self._spanRects shows the real coords for drawing:
(col, row) -> (x, y, width, height)
for each cell. Any cell which 'does not exist' as another
has spanned over it will get a None entry on the right.
This method generates relative positions so its results cannot
be reused after split.
"""
if getattr(self,'_spanRects',None): return
colpositions = self._colpositions
rowpositions = self._rowpositions
self._spanRects = spanRects = {}
self._vBlocks = vBlocks = {}
self._hBlocks = hBlocks = {}
for (coord, value) in self._spanRanges.items():
if value is None:
spanRects[coord] = None
else:
col,row = coord
# Testing row for visibility should be enough since no splits
# are permitted across spanned areas.
if not self._is_visible_row(row):
continue
col0, row0, col1, row1 = value
row0 = self._abs_to_vis(row0)
row1 = self._abs_to_vis(row1)
if col1-col0>0:
for _ in xrange(col0+1,col1+1):
vBlocks.setdefault(colpositions[_],[]).append((rowpositions[row1+1],rowpositions[row0]))
if row1-row0>0:
for _ in xrange(row0+1,row1+1):
hBlocks.setdefault(rowpositions[_],[]).append((colpositions[col0],colpositions[col1+1]))
x = colpositions[col0]
y = rowpositions[row1+1]
width = colpositions[col1+1] - x
height = rowpositions[row0] - y
spanRects[coord] = (x, y, width, height)
for _ in hBlocks, vBlocks:
for value in _.values():
value.sort()
def setStyle(self, tblstyle):
if not isinstance(tblstyle,TableStyle):
tblstyle = TableStyle(tblstyle)
for cmd in tblstyle.getCommands():
self._addCommand(cmd)
for k,v in tblstyle._opts.items():
setattr(self,k,v)
for a in ('spaceBefore','spaceAfter'):
if not hasattr(self,a) and hasattr(tblstyle,a):
setattr(self,a,getattr(tblstyle,a))
def _normalizeCoord(self, sc, ec, sr, er):
"""
Normalizes cols/rows coordinates.
"""
if sc < 0: sc = sc + self._ncols
if ec < 0: ec = ec + self._ncols
if sr < 0: sr = sr + self._nrows
if er < 0: er = er + self._nrows
return sc, ec, sr, er
def _addCommand(self,cmd):
if cmd[0] in ('BACKGROUND','ROWBACKGROUNDS','COLBACKGROUNDS'):
op, (sc, sr), (ec, er), arg = cmd
sc, ec, sr, er = self._normalizeCoord(sc, ec, sr, er)
cmd = (op, (sc, sr), (ec, er), arg)
self._bkgrndcmds.append(cmd)
elif cmd[0] == 'SPAN':
op, (sc, sr), (ec, er) = cmd
sc, ec, sr, er = self._normalizeCoord(sc, ec, sr, er)
if sc > ec: sc, ec = ec, sc
if sr > er: sr, er = er, sr
cmd = (op, (sc, sr), (ec, er))
self._spanCmds.append(cmd)
elif cmd[0] == 'NOSPLIT':
op, (sc, sr), (ec, er) = cmd
sc, ec, sr, er = self._normalizeCoord(sc, ec, sr, er)
if sc > ec: sc, ec = ec, sc
if sr > er: sr, er = er, sr
cmd = (op, (sc, sr), (ec, er))
self._nosplitCmds.append(cmd)
elif _isLineCommand(cmd):
# we expect op, start, stop, weight, colour, cap, dashes, join
cmd = list(cmd)
if len(cmd)<5: raise ValueError('bad line command '+str(cmd))
#determine line cap value at position 5. This can be str or numeric.
if len(cmd)<6:
cmd.append(1)
else:
cap = _convert2int(cmd[5], LINECAPS, 0, 2, 'cap', cmd)
cmd[5] = cap
#dashes at index 6 - this is a dash array:
if len(cmd)<7: cmd.append(None)
#join mode at index 7 - can be str or numeric, look up as for caps
if len(cmd)<8: cmd.append(1)
else:
join = _convert2int(cmd[7], LINEJOINS, 0, 2, 'join', cmd)
cmd[7] = join
#linecount at index 8. Default is 1, set to 2 for double line.
if len(cmd)<9: cmd.append(1)
else:
lineCount = cmd[8]
if lineCount is None:
lineCount = 1
cmd[8] = lineCount
assert lineCount >= 1
#linespacing at index 9. Not applicable unless 2+ lines, defaults to line
#width so you get a visible gap between centres
if len(cmd)<10: cmd.append(cmd[3])
else:
space = cmd[9]
if space is None:
space = cmd[3]
cmd[9] = space
assert len(cmd) == 10
(op, (sc,sr), (ec,er), weight, color, cap, dash, join, count,
space) = cmd[:]
sc, ec, sr, er = self._normalizeCoord(sc, ec, sr, er)
cmd = (op, (sc,sr), (ec,er), weight, color, cap, dash, join, count,
space)
self._linecmds.append(cmd)
else:
(op, (sc, sr), (ec, er)), values = cmd[:3] , cmd[3:]
sc, ec, sr, er = self._normalizeCoord(sc, ec, sr, er)
for i in xrange(sr, er+1):
for j in xrange(sc, ec+1):
_setCellStyle(self._cellStyles, i, j, op, values)
def _drawLines(self):
ccap, cdash, cjoin = None, None, None
self.canv.saveState()
for op, (sc,sr), (ec,er), weight, color, cap, dash, join, count, space in self._linecmds:
if isinstance(sr,basestring) and sr.startswith('split'): continue
if cap!=None and ccap!=cap:
self.canv.setLineCap(cap)
ccap = cap
if dash is None or dash == []:
if cdash is not None:
self.canv.setDash()
cdash = None
elif dash != cdash:
self.canv.setDash(dash)
cdash = dash
if join is not None and cjoin!=join:
self.canv.setLineJoin(join)
cjoin = join
getattr(self,_LineOpMap.get(op, '_drawUnknown' ))( (sc, sr), (ec, er), weight, color, count, space)
self.canv.restoreState()
self._curcolor = None
def _drawUnknown(self, (sc, sr), (ec, er), weight, color, count, space):
#we are only called from _drawLines which is one level up
import sys
op = sys._getframe(1).f_locals['op']
raise ValueError("Unknown line command '%s'" % op)
def _is_visible_line(self, line_num):
"""
Checks if line is in visible area of table.
"""
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
return (line_num <= self.repeatRows or
activeRows0 <= line_num <= self._activeRows[1] or
self._nrows - self._repeatRowsB <= line_num <= self._nrows)
def _is_visible_row(self, row_num):
"""
Checks if row is in visible area of table.
"""
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
return (row_num < self.repeatRows or
activeRows0 <= row_num < self._activeRows[1] or
self._nrows - self._repeatRowsB <= row_num < self._nrows)
def _abs_to_vis(self, line_num):
"""
Translates absolute line positions to relative (visible positions).
"""
if line_num <= self.repeatRows:
return line_num
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
if activeRows0 <= line_num <= self._activeRows[1]:
line_num -= activeRows0 - self.repeatRows
return line_num
if self._nrows - self._repeatRowsB <= line_num <= self._nrows:
line_num -= activeRows0 - self.repeatRows
line_num -= self._nrows - self._repeatRowsB - self._activeRows[1]
return line_num
raise ValueError('`line_num` outside visible area!')
def _vis_to_abs(self, line_num):
"""
Translates relative line positions to absolute.
"""
if line_num <= self.repeatRows:
return line_num
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
line_num += activeRows0 - self.repeatRows
if activeRows0 <= line_num <= self._activeRows[1]:
return line_num
line_num += self._nrows - self._repeatRowsB - self._activeRows[1]
if self._nrows - self._repeatRowsB <= line_num <= self._nrows:
return line_num
raise ValueError('`line_num` outside visible area!')
def _drawGrid(self, (sc, sr), (ec, er), weight, color, count, space):
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
# Checks if whole grid is outside visible area.
if sr >= self.repeatRows and er < activeRows0:
return
if sr >= self._activeRows[1] and er < self._nrows - self._repeatRowsB:
return
# Some parts visible - rendering.
self._drawBox( (sc, sr), (ec, er), weight, color, count, space)
self._drawInnerGrid( (sc, sr), (ec, er), weight, color, count, space)
def _drawBox(self, (sc, sr), (ec, er), weight, color, count, space):
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
# Checks if whole box is outside visible area.
if sr >= self.repeatRows and er < activeRows0:
return
if sr >= self._activeRows[1] and er < self._nrows - self._repeatRowsB:
return
# Some parts visible - rendering.
# If start row visible renders upper horizontal line.
if self._is_visible_row(sr):
self._drawHLines((sc, sr), (ec, sr), weight, color, count, space)
# If end row visible renders lower horizontal line.
if self._is_visible_row(er):
self._drawHLines((sc, er+1), (ec, er+1), weight, color, count, space)
# Renders vertical lines.
self._drawVLines((sc, sr), (sc, er + 1), weight, color, count, space)
self._drawVLines((ec+1, sr), (ec+1, er + 1), weight, color, count, space)
def _drawInnerGrid(self, (sc, sr), (ec, er), weight, color, count, space):
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
# Checks if whole inner grid is outside visible area.
if sr >= self.repeatRows and er < activeRows0:
return
if sr >= self._activeRows[1] and er < self._nrows - self._repeatRowsB:
return
# Some parts visible - rendering.
self._drawHLines((sc, sr+1), (ec, er), weight, color, count, space)
self._drawVLines((sc+1, sr), (ec, er + 1), weight, color, count, space)
def _drawLineAbove(self, (sc, sr), (ec, er), weight, color, count, space):
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
# Checks if whole row range is outside visible area.
if sr >= self.repeatRows and er < activeRows0:
return
if sr >= self._activeRows[1] and er < self._nrows - self._repeatRowsB:
return
# Some parts visible - searching for visible rows.
visible = []
for i in xrange(sr, er + 1):
if not self._is_visible_row(i):
continue
visible.append(i)
# Generates line for each visible row.
for vis in visible:
self._drawHLines((sc, vis), (ec, vis), weight, color, count, space)
def _drawLineBelow(self, (sc, sr), (ec, er), weight, color, count, space):
activeRows0 = self._activeRows[0] if self._activeRows[0] is not None else self.repeatRows # ugly hack to make it backward compatible
# Checks if whole row range is outside visible area.
if sr >= self.repeatRows and er < activeRows0:
return
if sr >= self._activeRows[1] and er < self._nrows - self._repeatRowsB:
return
# Some parts visible - searching for visible rows.
visible = []
for i in xrange(sr, er + 1):
if not self._is_visible_row(i):
continue
visible.append(i)
# Generates line for each visible row.
for vis in visible:
self._drawHLines((sc, vis + 1), (ec, vis +1), weight, color, count, space)
def _prepLine(self, weight, color):
if color != self._curcolor:
self.canv.setStrokeColor(color)
self._curcolor = color
if weight != self._curweight:
self.canv.setLineWidth(weight)