B \'@sdZddlZddlZddlZddlZddlZddlZddlZddlm Z m Z e rXej Z nej Z Gddde ZGddde ZGd d d e ZGd d d eZdddZdS)zb Fast cryptographic hash of Python objects, with a special case for fast hashing of numpy arrays. N)_bytes_or_unicode PY3_OR_LATERc@seZdZdZddZdS)_ConsistentSetza Class used to ensure the hash of Sets is preserved whatever the order of its items. c CsByt||_Wn.ttjfk r<tdd|D|_YnXdS)Ncss|]}t|VqdS)N)hash).0er ?lib/python3.7/site-packages/sklearn/externals/joblib/hashing.py *sz*_ConsistentSet.__init__..)sortedZ _sequence TypeErrordecimalZInvalidOperation)selfZ set_sequencer r r __init__sz_ConsistentSet.__init__N)__name__ __module__ __qualname____doc__rr r r r rsrc@seZdZdZddZdS)_MyHashz7 Class used to hash objects that won't normally pickle cGs ||_dS)N)args)rrr r r r0sz_MyHash.__init__N)rrrrrr r r r r-src@seZdZdZdddZdddZdd Zd d Zd ej fd dZ e j Z e e ee<e e ee<e e ee <e e eej<ddZddZee ee<d S)HasherzW A subclass of pickler, to do cryptographic hashing, rather than pickling. md5cCs<t|_trtjntj}tj||j|dt ||_ dS)N)protocol) ioBytesIOstreamrpickleZDEFAULT_PROTOCOLZHIGHEST_PROTOCOLPicklerrhashlibnew_hash)r hash_namerr r r r9s   zHasher.__init__Tc Csvy||Wn>tjk rL}z|jd||ff7_Wdd}~XYnX|j}|j||rr|jSdS)Nz"PicklingError while hashing %r: %r) dumpr PicklingErrorrrgetvaluer!updateZ hexdigest)robjZ return_digestrdumpsr r r rCs  z Hasher.hashcCst|tjtijfr|t|dr*|jj}n|j}|j}t|tt krTt ||j}n(|dkrht ||}n|jj }t |||}t ||dS)N__func__) isinstancetypes MethodTypetypepophasattrr)r__self__rr __class__rsave)rr'Z func_nameZinstclsr r r r2Ns    z Hasher.savecCst|trdSt||dS)N)r*rrmemoize)rr'r r r r4as zHasher.memoizeNcCst||d}tjdkr|d=ytj||f|Wnltjk rtj||f|t|dd}|dkr|}|dkrx|j}tj |}t ||st |||YnXdS)N)namepack)r6r__main__) dictsys version_infor save_globalrr$getattrrmodulesr/setattr)rr'r5r6kwargsmoduleZmy_namemodr r r r=ls     zHasher.save_globalc CsPyt|tt|Wn2tk rJt|ttdd|DYnXdS)Ncss|]\}}t||fVqdS)N)r)rkvr r r r sz)Hasher._batch_setitems..)r_batch_setitemsiterr r )ritemsr r r rFs zHasher._batch_setitemscCst|t|dS)N)rr2r)rZ set_itemsr r r save_setszHasher.save_set)r)T)rrrrrrr2r4structr6r=rdispatchcopyr-lenobjectrr#rFrIsetr r r r r4s      rc@s"eZdZdZd ddZddZdS) NumpyHasherz7 Special case the hasher for when numpy is loaded. rFcCs@||_tj||dddl}||_t|dr6|j|_nt|_dS)a Parameters ---------- hash_name: string The hash algorithm to be used coerce_mmap: boolean Make no difference between np.memmap and np.ndarray objects. )r"rN getbuffer) coerce_mmaprrnumpynpr/rQ _getbuffer memoryview)rr"rRrTr r r rs   zNumpyHasher.__init__cCst||jjr|jjs|jdkr*|}n&|jjr8|}n|jj rH|j }n|}|j | ||jj|jrt||jjr|jj}n|j}|d|j|j|jff}n"t||jjr|j}|d|jff}t||dS)z Subclass the save method, to hash ndarray subclass, rather than pickling them. Off course, this is a total abuse of the Pickler class. r ZHASHEDN)r*rTZndarrayZdtypeZ hasobjectshapeZflattenflags c_contiguous f_contiguousTr!r&rUZviewZuint8rRZmemmapr1stridesZdescrrr2)rr'Zobj_c_contiguousklassr r r r2s$    zNumpyHasher.saveN)rF)rrrrrr2r r r r rPs rPrFcCs,dtjkrt||d}n t|d}||S)al Quick calculation of a hash to identify uniquely Python objects containing numpy arrays. Parameters ----------- hash_name: 'md5' or 'sha1' Hashing algorithm used. sha1 is supposedly safer, but md5 is faster. coerce_mmap: boolean Make no difference between np.memmap and np.ndarray rS)r"rR)r")r;r?rPrr)r'r"rRZhasherr r r rs  r)rF)rrrr;r+rJrrZ_compatrrZ_PicklerrrNrrrrPrr r r r s mU