i8 %SSKrSSKrSSKrSSKrSSKrSSKJrJr SSKJ r J r SSK J r SSK Jr SSKJrJr SSKJr SSKJr SS KJr SS KJr \(aSS KJr S rS \S\4SjrS\ S\4Sjr S\!S\4Sjr"S\#S\4Sjr$S\S\4Sjr%\"S5r&\"S5r'\"S5r(S\(S\(4Sjr)S\!\RT\ \(-S\S\ \(4Sjr+\ "S 5S!\,\-S"4S\.\,\\\-44S#j5r/"S$S%5r0S&r1\0"S'S(S)9r2"S*S+5r3\"S,5=(d \RhRjr6\-\7S-'\ S.\-S/\\--S\-4S0j5r8S1\-S/\\--S\-4S2jr9g)3N) OrderedDictabc)CallableSequence) lru_cache) FunctionType) TYPE_CHECKINGTypeVar)InvalidArgument)int_from_bytes)next_up) _function_key)ConjectureDatalnamereturnc[R"UR55R5n[ USS5$)N)hashlibsha384encodedigestr )rhasheds c/var/www/html/ai-image-ml/venv/lib/python3.13/site-packages/hypothesis/internal/conjecture/utils.pycalc_label_from_namer"s0 ^^DKKM * 1 1 3F &!* %%fc[U[5(a[[USS95$[U[5(a [ U5$[U5$![ a [ [ U55s$f=f)NT) ignore_name) isinstancercalc_label_from_hashrtypecalc_label_from_cls Exception)rs rcalc_label_from_callabler$'sg!\""#M!$FGG At  "1%% 0'* * 0&tAw/ / 0s AA65A6clsc,[UR5$N)r __qualname__)r%s rr"r"5s  0 0 11robjc<[[[U555$r')rstrhash)r)s rr r 9s DI //rlabelsc<SnUHnUS-[-nX-nM U$)Nr) LABEL_MASK)r-labells rcombine_labelsr3=s- E !z)  Lrza sample() in Samplerzone more from many()TvcU$r')r5s ridentityr8Ls Hrvalues strategy_namecS[R;ac[U[RSR5(a7URS:wa&[ SURSUR S35eOG[U[[R[R45(d[ SU<SUS35e[U[5(aU$[U5$) Nnumpyr/zPOnly one-dimensional arrays are supported for sampling, and the given value has z dimensions (shape z). This array would give samples of array slices instead of elements! Use np.ravel(values) to convert to a one-dimensional array, or tuple(values) if you want to sample slices.zCannot sample from z\ because it is not an ordered collection. Hypothesis goes to some length to ensure that the a strategy has stable results between runs. To replay a saved example, the sampled values must have the same iteration order on every run - ruling out sets, dicts, etc due to hash randomization. Most cases can simply use `sorted(values)`, but mixed types or special values such as math.nan require careful handling - and note that when simplifying an example, Hypothesis treats earlier values as simpler.)sysmodulesrndarrayndimr shaperrrenumEnumMetarangetuple)r9r:s r check_samplerFPs#++*VS[[5I5Q5Q"R"R ;;! !++1;;-7J<<.!))  cllDMM J K K!&,AANP; ;  &%   =r@weights.c[U5n[U5Vs/sHo"SS/PM nn[U5n[U5nU"S5nU"S5n/n/n UV s/sHoU- PM n n /n [ U 5HSup-X-nU R U5 US:Xa XcUS'M)US:aUR U5 MBU R U5 MU [ R"U5 [ R"U 5 U(aU (a[ R"U5n[ R"U 5nUU:wdeU UU:deX?SbeUX?S'X|U- X?S'U UX-U- U U'U US:a[ R"UU5 O(U US:XaXcUS'O[ R"U U5 U(a U (aMU (aXcU R5S'U (aMU(aXcUR5S'U(aM/nUHunnn [U[5(de[U[5(dUbeU ceUcUR UUU 45 MVUU:aUR UUX}- 45 MtUR UUU 45 M UR5 U$s snfs sn f)Nrr/)lenrDsumr! enumerateappendheapqheapifyheappopheappushpoprintsort)rHnitabletotalnum_typezeroonesmalllargew probabilitiesscaled_probabilitiesalternate_chancescaledlohi new_tablebase alternates rcompute_sampler_tablerips G AFKAh,Oht_hE,O LEE{H A;D 1+CEE(/01YM0(*(7!%##F+ Q;!HQK aZ LLO LLO 8 MM% MM% E ]]5 ! ]]5 !Rxx#B'#---y|### ! "55 !  $';'? ? $R   #a ' NN5" % !" % *"IaL NN5" %% EE&  $eiik1 %  $eiik1 %/1I-2)i)$$$$$)S))Y->>>+++     dD*:; <     is/EF G   dI/?@ A.3NN {-P1s J6J;cz\rSrSr%Sr\\\\\4\ S'SS.S\ \S\ 4Sjjr S S .S S S \S -S\4Sjjr Srg )SampleraSampler based on Vose's algorithm for the alias method. See http://www.keithschwarz.com/darts-dice-coins/ for a good explanation. The general idea is that we store a table of triples (base, alternate, p). base. We then pick a triple uniformly at random, and choose its alternate value with probability p and else choose its base value. The triples are chosen so that the resulting mixture has the right distribution. We maintain the following invariants to try to produce good shrinks: 1. The table is in lexicographic (base, alternate) order, so that choosing an earlier value in the list always lowers (or at least leaves unchanged) the value. 2. base[i] < alternate[i], so that shrinking the draw always results in shrinking the chosen element. rXTobserverHrncBX l[[U55Ulgr')rnrirErX)selfrHrns r__init__Sampler.__init__s *5>: rN)forceddatarrsrc^UR(aUR[5 TcSO[U4SjUR55nUR URUURS9upEnSnTb TU:H=(a US:nTU:Xd U(deUR UUURS9nUR(aUR5 U(aTbUT:Xd TU45eU$TbUT:Xd TU45eU$)Nc3^># UH"upnTU:XdTU:XdMUS:dMXU4v M$ g7f)rNr7).0rgrhrbrss r !Sampler.sample..sB;E7T&6T>f &94>NQR>R4"23;Es -- -rsrnr)rn start_spanSAMPLE_IN_SAMPLER_LABELnextrXchoice draw_boolean stop_span) rprtrs forced_choicergrhrbforced_use_alternate use_alternates ` rsampleSampler.samples << OO3 4~ ;?:: -1KK JJ LL-8- )) $  $*Y#6#O;Ka;O T>%9 99)) 'LL* << NN  >Y&%8 M69:M M8 >TV^ Cfd^ C3Kr)rnrX)__name__ __module__r(__firstlineno____doc__listrErTfloat__annotations__rboolrqr__static_attributes__r7rrrkrksh" c3o& ''DH;;T;" **d * **rrk)r rG)g@g @?rg?Frmc\rSrSrSrSSS.SSS\S \\-S \\-S \S-S \S S4SjjrSr Sr S \4Sjr SS\ S-S S4Sjjr Srg)manyaUtility class for collections. Bundles up the logic we use for "should I keep drawing more values?" and handles starting and stopping examples in the right place. Intended usage is something like: elements = many(data, ...) while elements.more(): add_stuff_to_result() NTrzrtrmin_sizemax_size average_sizersrnrc SUs=::aUs=::aU::de eUbX%s=::aU::de eX lX0lXlXPl[ XB- X2- 5UlSUlSUlSUlSUl SUl X`l g)NrF) rrrt forced_size_calc_p_continue p_continuecount rejectionsdrawn force_stoprejectedrn)rprtrrrrsrns rrq many.__init__sH8 888888~!?x!???!???   !*<+BHDWX    rc\UR(aURR5 ggr')rnrtr)rps rrmany.stop_span s << II   ! rc^UR(aURRU5 ggr')rnrtr{)rpr1s rr{many.start_span$s << II  ' rcUR(aUR5 SUlSUlUR[5 UR UR :XaURUR :nOSnUR(a,URbURUR:XdeSnO`URUR :aSnOCURUR :aSnO&URbURUR:nURRURUURS9nU(aU=RS- slgUR5 g)z7Should I draw another element to add to the collection?TFNrzr/)rrrr{ONE_FROM_MANY_LABELrrrrrrtrrrn)rpshould_continue forced_results rmore many.more(s :: NN    +, ==DMM )#jj4==8O M''/4::AQAQ3QQQ % dmm+ $ t}}, % !!- $ T-=-= = "ii44$ 5O  JJ!OJ NN rwhycTURS:deU=RS-slU=RS- slSUlUR[SSUR-5:a>URUR:aUR R U5 gSUlgg)zReject the last example (i.e. don't count it towards our budget of elements because it's not going to go in the final collection).rr/TrJN)rrrmaxrrt mark_invalidr)rprs rreject many.rejectOszzA~~ a  1  ??SA N3 3zzDMM) &&s+"& 4r) rrtrrrrrrnrrrr')rrr(rrrTrrrqrr{rr+rrr7rrrrs $"+  Ek d  0"(%d%N '#* ' ' 'rrgSMALLEST_POSITIVE_FLOAT desired_avgrcdX::dX45eX:XagSSSU-- - nUS:XdU[R:XaSUs=::a S:dU5e U5eU$SUs=:a S:dU5e U5e[X!5U:a'US-nU[:a[nO[X!5U:aM'SnU[X!5- S:aFSUs=:aU:d O X#45eX#-S- n[XA5U::aUnOUnU[X!5- S:aMFSUs=:a S:dU5e U5e[X!5U::deU$)zCReturn the p_continue which will generate the desired average size.rr/rg-C6?g{Gz?rJ)mathinf_p_continue_to_avgr)rrrremids rrrbsx  ";[$;; "SA O,,JQ(dhh.J"".J.".J." z A )z) )z)  Z 2[ @f / /0J  Z 2[ @ B *:@ @4 G:""4Z$44"!# c , ;JB *:@ @4 G z A )z) )z)  j 3{ BB B rrc8US:aU$SSU- - S- SX-- -$)zBReturn the average_size generated by this p_continue and max_size.r/rr7)rrs rrrs/Q 1z> "Q &1z/C+C DDr):rBrrOrr= collectionsrrcollections.abcrr functoolsrtypesrtypingr r hypothesis.errorsr hypothesis.internal.compatr hypothesis.internal.floatsr "hypothesis.internal.lambda_sourcesr#hypothesis.internal.conjecture.datarr0r+rTrr$r!r"objectr r3r|rr4r8EnumrFrErrrirk INT_SIZESINT_SIZES_SAMPLERr float_infominrrrrr7rrrs  (.)-5.<B &s&s& 0 0S 02T2c20f00CC//FG*+AB CL  a  Ohqk ):= a[@ 2?5#4?eCeO>T9U??DBBJ ! 5uE`'`'F")!C1C1CC #%#3;#5# #LE5EC%KEEEr