pyerrors.input.dobs

  1from collections import defaultdict
  2import gzip
  3import lxml.etree as et
  4import getpass
  5import socket
  6import datetime
  7import json
  8import warnings
  9import numpy as np
 10from ..obs import Obs
 11from ..obs import _merge_idx
 12from ..covobs import Covobs
 13from .. import version as pyerrorsversion
 14
 15
 16# Based on https://stackoverflow.com/a/10076823
 17def _etree_to_dict(t):
 18    """ Convert the content of an XML file to a python dict"""
 19    d = {t.tag: {} if t.attrib else None}
 20    children = list(t)
 21    if children:
 22        dd = defaultdict(list)
 23        for dc in map(_etree_to_dict, children):
 24            for k, v in dc.items():
 25                dd[k].append(v)
 26        d = {t.tag: {k: v[0] if len(v) == 1 else v
 27                     for k, v in dd.items()}}
 28    if t.attrib:
 29        d[t.tag].update(('@' + k, v)
 30                        for k, v in t.attrib.items())
 31    if t.text:
 32        text = t.text.strip()
 33        if children or t.attrib:
 34            if text:
 35                d[t.tag]['#data'] = [text]
 36        else:
 37            d[t.tag] = text
 38    return d
 39
 40
 41def _dict_to_xmlstring(d):
 42    if isinstance(d, dict):
 43        iters = ''
 44        for k in d:
 45            if k.startswith('#'):
 46                for la in d[k]:
 47                    iters += la
 48                iters = '<array>\n' + iters + '<%sarray>\n' % ('/')
 49                return iters
 50            if isinstance(d[k], dict):
 51                iters += '<%s>\n' % (k) + _dict_to_xmlstring(d[k]) + '<%s%s>\n' % ('/', k)
 52            elif isinstance(d[k], str):
 53                if len(d[k]) > 100:
 54                    iters += '<%s>\n ' % (k) + d[k] + ' \n<%s%s>\n' % ('/', k)
 55                else:
 56                    iters += '<%s> ' % (k) + d[k] + ' <%s%s>\n' % ('/', k)
 57            elif isinstance(d[k], list):
 58                for i in range(len(d[k])):
 59                    iters += _dict_to_xmlstring(d[k][i])
 60            elif not d[k]:
 61                return '\n'
 62            else:
 63                raise Exception('Type', type(d[k]), 'not supported in export!')
 64    else:
 65        raise Exception('Type', type(d), 'not supported in export!')
 66    return iters
 67
 68
 69def _dict_to_xmlstring_spaces(d, space='  '):
 70    s = _dict_to_xmlstring(d)
 71    o = ''
 72    c = 0
 73    cm = False
 74    for li in s.split('\n'):
 75        if li.startswith('<%s' % ('/')):
 76            c -= 1
 77            cm = True
 78        for i in range(c):
 79            o += space
 80        o += li + '\n'
 81        if li.startswith('<') and not cm:
 82            if not '<%s' % ('/') in li:
 83                c += 1
 84        cm = False
 85    return o
 86
 87
 88def create_pobs_string(obsl, name, spec='', origin='', symbol=[], enstag=None):
 89    """Export a list of Obs or structures containing Obs to an xml string
 90    according to the Zeuthen pobs format.
 91
 92    Tags are not written or recovered automatically. The separator | is removed from the replica names.
 93
 94    Parameters
 95    ----------
 96    obsl : list
 97        List of Obs that will be exported.
 98        The Obs inside a structure have to be defined on the same ensemble.
 99    name : str
100        The name of the observable.
101    spec : str
102        Optional string that describes the contents of the file.
103    origin : str
104        Specify where the data has its origin.
105    symbol : list
106        A list of symbols that describe the observables to be written. May be empty.
107    enstag : str
108        Enstag that is written to pobs. If None, the ensemble name is used.
109    """
110
111    od = {}
112    ename = obsl[0].e_names[0]
113    names = list(obsl[0].deltas.keys())
114    nr = len(names)
115    onames = [name.replace('|', '') for name in names]
116    for o in obsl:
117        if len(o.e_names) != 1:
118            raise Exception('You try to export dobs to obs!')
119        if o.e_names[0] != ename:
120            raise Exception('You try to export dobs to obs!')
121        if len(o.deltas.keys()) != nr:
122            raise Exception('Incompatible obses in list')
123    od['observables'] = {}
124    od['observables']['schema'] = {'name': 'lattobs', 'version': '1.0'}
125    od['observables']['origin'] = {
126        'who': getpass.getuser(),
127        'date': str(datetime.datetime.now())[:-7],
128        'host': socket.gethostname(),
129        'tool': {'name': 'pyerrors', 'version': pyerrorsversion.__version__}}
130    od['observables']['pobs'] = {}
131    pd = od['observables']['pobs']
132    pd['spec'] = spec
133    pd['origin'] = origin
134    pd['name'] = name
135    if enstag:
136        if not isinstance(enstag, str):
137            raise Exception('enstag has to be a string!')
138        pd['enstag'] = enstag
139    else:
140        pd['enstag'] = ename
141    pd['nr'] = '%d' % (nr)
142    pd['array'] = []
143    osymbol = 'cfg'
144    if not isinstance(symbol, list):
145        raise Exception('Symbol has to be a list!')
146    if not (len(symbol) == 0 or len(symbol) == len(obsl)):
147        raise Exception('Symbol has to be a list of lenght 0 or %d!' % (len(obsl)))
148    for s in symbol:
149        osymbol += ' %s' % s
150    for r in range(nr):
151        ad = {}
152        ad['id'] = onames[r]
153        Nconf = len(obsl[0].deltas[names[r]])
154        layout = '%d i f%d' % (Nconf, len(obsl))
155        ad['layout'] = layout
156        ad['symbol'] = osymbol
157        data = ''
158        for c in range(Nconf):
159            data += '%d ' % obsl[0].idl[names[r]][c]
160            for o in obsl:
161                num = o.deltas[names[r]][c] + o.r_values[names[r]]
162                if num == 0:
163                    data += '0 '
164                else:
165                    data += '%1.16e ' % (num)
166            data += '\n'
167        ad['#data'] = data
168        pd['array'].append(ad)
169
170    rs = '<?xml version="1.0" encoding="utf-8"?>\n' + _dict_to_xmlstring_spaces(od)
171    return rs
172
173
174def write_pobs(obsl, fname, name, spec='', origin='', symbol=[], enstag=None, gz=True):
175    """Export a list of Obs or structures containing Obs to a .xml.gz file
176    according to the Zeuthen pobs format.
177
178    Tags are not written or recovered automatically. The separator | is removed from the replica names.
179
180    Parameters
181    ----------
182    obsl : list
183        List of Obs that will be exported.
184        The Obs inside a structure have to be defined on the same ensemble.
185    fname : str
186        Filename of the output file.
187    name : str
188        The name of the observable.
189    spec : str
190        Optional string that describes the contents of the file.
191    origin : str
192        Specify where the data has its origin.
193    symbol : list
194        A list of symbols that describe the observables to be written. May be empty.
195    enstag : str
196        Enstag that is written to pobs. If None, the ensemble name is used.
197    gz : bool
198        If True, the output is a gzipped xml. If False, the output is an xml file.
199    """
200    pobsstring = create_pobs_string(obsl, name, spec, origin, symbol, enstag)
201
202    if not fname.endswith('.xml') and not fname.endswith('.gz'):
203        fname += '.xml'
204
205    if gz:
206        if not fname.endswith('.gz'):
207            fname += '.gz'
208
209        fp = gzip.open(fname, 'wb')
210        fp.write(pobsstring.encode('utf-8'))
211    else:
212        fp = open(fname, 'w', encoding='utf-8')
213        fp.write(pobsstring)
214    fp.close()
215
216
217def _import_data(string):
218    return json.loads("[" + ",".join(string.replace(' +', ' ').split()) + "]")
219
220
221def _check(condition):
222    if not condition:
223        raise Exception("XML file format not supported")
224
225
226class _NoTagInDataError(Exception):
227    """Raised when tag is not in data"""
228    def __init__(self, tag):
229        self.tag = tag
230        super().__init__('Tag %s not in data!' % (self.tag))
231
232
233def _find_tag(dat, tag):
234    for i in range(len(dat)):
235        if dat[i].tag == tag:
236            return i
237    raise _NoTagInDataError(tag)
238
239
240def _import_array(arr):
241    name = arr[_find_tag(arr, 'id')].text.strip()
242    index = _find_tag(arr, 'layout')
243    try:
244        sindex = _find_tag(arr, 'symbol')
245    except _NoTagInDataError:
246        sindex = 0
247    if sindex > index:
248        tmp = _import_data(arr[sindex].tail)
249    else:
250        tmp = _import_data(arr[index].tail)
251
252    li = arr[index].text.strip()
253    m = li.split()
254    if m[1] == "i" and m[2][0] == "f":
255        nc = int(m[0])
256        na = int(m[2].lstrip('f'))
257        _dat = []
258        mask = []
259        for a in range(na):
260            mask += [a]
261            _dat += [np.array(tmp[1 + a:: na + 1])]
262        _check(len(tmp[0:: na + 1]) == nc)
263        return [name, tmp[0:: na + 1], mask, _dat]
264    elif m[1][0] == 'f' and len(m) < 3:
265        sh = (int(m[0]), int(m[1].lstrip('f')))
266        return np.reshape(tmp, sh)
267    elif any(['f' in s for s in m]):
268        for si in range(len(m)):
269            if m[si] == 'f':
270                break
271        sh = [int(m[i]) for i in range(si)]
272        return np.reshape(tmp, sh)
273    else:
274        print(name, m)
275        _check(False)
276
277
278def _import_rdata(rd):
279    name, idx, mask, deltas = _import_array(rd)
280    return deltas, name, idx
281
282
283def _import_cdata(cd):
284    _check(cd[0].tag == "id")
285    _check(cd[1][0].text.strip() == "cov")
286    cov = _import_array(cd[1])
287    grad = _import_array(cd[2])
288    return cd[0].text.strip(), cov, grad
289
290
291def read_pobs(fname, full_output=False, gz=True, separator_insertion=None):
292    """Import a list of Obs from an xml.gz file in the Zeuthen pobs format.
293
294    Tags are not written or recovered automatically.
295
296    Parameters
297    ----------
298    fname : str
299        Filename of the input file.
300    full_output : bool
301        If True, a dict containing auxiliary information and the data is returned.
302        If False, only the data is returned as list.
303    separatior_insertion: str or int
304        str: replace all occurences of "separator_insertion" within the replica names
305        by "|%s" % (separator_insertion) when constructing the names of the replica.
306        int: Insert the separator "|" at the position given by separator_insertion.
307        None (default): Replica names remain unchanged.
308    """
309
310    if not fname.endswith('.xml') and not fname.endswith('.gz'):
311        fname += '.xml'
312    if gz:
313        if not fname.endswith('.gz'):
314            fname += '.gz'
315        with gzip.open(fname, 'r') as fin:
316            content = fin.read()
317    else:
318        if fname.endswith('.gz'):
319            warnings.warn("Trying to read from %s without unzipping!" % fname, UserWarning)
320        with open(fname, 'r') as fin:
321            content = fin.read()
322
323    # parse xml file content
324    root = et.fromstring(content)
325
326    _check(root[2].tag == 'pobs')
327    pobs = root[2]
328
329    version = root[0][1].text.strip()
330
331    _check(root[1].tag == 'origin')
332    file_origin = _etree_to_dict(root[1])['origin']
333
334    deltas = []
335    names = []
336    idl = []
337    for i in range(5, len(pobs)):
338        delta, name, idx = _import_rdata(pobs[i])
339        deltas.append(delta)
340        if separator_insertion is None:
341            pass
342        elif isinstance(separator_insertion, int):
343            name = name[:separator_insertion] + '|' + name[separator_insertion:]
344        elif isinstance(separator_insertion, str):
345            name = name.replace(separator_insertion, "|%s" % (separator_insertion))
346        else:
347            raise Exception("separator_insertion has to be string or int, is ", type(separator_insertion))
348        names.append(name)
349        idl.append(idx)
350    res = [Obs([d[i] for d in deltas], names, idl=idl) for i in range(len(deltas[0]))]
351
352    descriptiond = {}
353    for i in range(4):
354        descriptiond[pobs[i].tag] = pobs[i].text.strip()
355
356    _check(pobs[4].tag == "nr")
357
358    _check(pobs[5].tag == 'array')
359    if pobs[5][1].tag == 'symbol':
360        symbol = pobs[5][1].text.strip()
361        descriptiond['symbol'] = symbol
362
363    if full_output:
364        retd = {}
365        tool = file_origin.get('tool', None)
366        if tool:
367            program = tool['name'] + ' ' + tool['version']
368        else:
369            program = ''
370        retd['program'] = program
371        retd['version'] = version
372        retd['who'] = file_origin['who']
373        retd['date'] = file_origin['date']
374        retd['host'] = file_origin['host']
375        retd['description'] = descriptiond
376        retd['obsdata'] = res
377        return retd
378    else:
379        return res
380
381
382# Reading (and writing) dobs is not yet working properly:
383# we have to loop over root[2:] because each entry is a dobs
384# But maybe this is just a problem with Ben's implementation
385
386# this is based on Mattia Bruno's implementation at https://github.com/mbruno46/pyobs/blob/master/pyobs/IO/xml.py
387def import_dobs_string(content, noempty=False, full_output=False, separator_insertion=True):
388    """Import a list of Obs from a string in the Zeuthen dobs format.
389
390    Tags are not written or recovered automatically.
391
392    Parameters
393    ----------
394    content : str
395        XML string containing the data
396    noemtpy : bool
397        If True, ensembles with no contribution to the Obs are not included.
398        If False, ensembles are included as written in the file, possibly with vanishing entries.
399    full_output : bool
400        If True, a dict containing auxiliary information and the data is returned.
401        If False, only the data is returned as list.
402    separatior_insertion: str, int or bool
403        str: replace all occurences of "separator_insertion" within the replica names
404        by "|%s" % (separator_insertion) when constructing the names of the replica.
405        int: Insert the separator "|" at the position given by separator_insertion.
406        True (default): separator "|" is inserted after len(ensname), assuming that the
407        ensemble name is a prefix to the replica name.
408        None or False: No separator is inserted.
409    """
410
411    root = et.fromstring(content)
412
413    _check(root.tag == 'OBSERVABLES')
414    _check(root[0].tag == 'SCHEMA')
415    version = root[0][1].text.strip()
416
417    _check(root[1].tag == 'origin')
418    file_origin = _etree_to_dict(root[1])['origin']
419
420    _check(root[2].tag == 'dobs')
421
422    dobs = root[2]
423
424    descriptiond = {}
425    for i in range(3):
426        descriptiond[dobs[i].tag] = dobs[i].text.strip()
427
428    _check(dobs[3].tag == 'array')
429
430    symbol = []
431    if dobs[3][1].tag == 'symbol':
432        symbol = dobs[3][1].text.strip()
433        descriptiond['symbol'] = symbol
434    mean = _import_array(dobs[3])[0]
435
436    _check(dobs[4].tag == "ne")
437    ne = int(dobs[4].text.strip())
438    _check(dobs[5].tag == "nc")
439    nc = int(dobs[5].text.strip())
440
441    idld = {}
442    deltad = {}
443    covd = {}
444    gradd = {}
445    names = []
446    e_names = []
447    enstags = {}
448    for k in range(6, len(list(dobs))):
449        if dobs[k].tag == "edata":
450            _check(dobs[k][0].tag == "enstag")
451            ename = dobs[k][0].text.strip()
452            e_names.append(ename)
453            _check(dobs[k][1].tag == "nr")
454            R = int(dobs[k][1].text.strip())
455            for i in range(2, 2 + R):
456                deltas, rname, idx = _import_rdata(dobs[k][i])
457                if separator_insertion is None or False:
458                    pass
459                elif separator_insertion is True:
460                    if rname.startswith(ename):
461                        rname = rname[:len(ename)] + '|' + rname[len(ename):]
462                elif isinstance(separator_insertion, int):
463                    rname = rname[:separator_insertion] + '|' + rname[separator_insertion:]
464                elif isinstance(separator_insertion, str):
465                    rname = rname.replace(separator_insertion, "|%s" % (separator_insertion))
466                else:
467                    raise Exception("separator_insertion has to be string or int, is ", type(separator_insertion))
468                if '|' in rname:
469                    new_ename = rname[:rname.index('|')]
470                else:
471                    new_ename = ename
472                enstags[new_ename] = ename
473                idld[rname] = idx
474                deltad[rname] = deltas
475                names.append(rname)
476        elif dobs[k].tag == "cdata":
477            cname, cov, grad = _import_cdata(dobs[k])
478            covd[cname] = cov
479            if grad.shape[1] == 1:
480                gradd[cname] = [grad for i in range(len(mean))]
481            else:
482                gradd[cname] = grad.T
483        else:
484            _check(False)
485    names = list(set(names))
486
487    for name in names:
488        for i in range(len(deltad[name])):
489            deltad[name][i] = np.array(deltad[name][i]) + mean[i]
490
491    res = []
492    for i in range(len(mean)):
493        deltas = []
494        idl = []
495        obs_names = []
496        for name in names:
497            h = np.unique(deltad[name][i])
498            if len(h) == 1 and np.all(h == mean[i]) and noempty:
499                continue
500            deltas.append(deltad[name][i])
501            obs_names.append(name)
502            idl.append(idld[name])
503        res.append(Obs(deltas, obs_names, idl=idl))
504    print(mean, 'vs', res)
505    _check(len(e_names) == ne)
506
507    cnames = list(covd.keys())
508    for i in range(len(res)):
509        new_covobs = {name: Covobs(0, covd[name], name, grad=gradd[name][i]) for name in cnames}
510        if noempty:
511            for name in cnames:
512                if np.all(new_covobs[name].grad == 0):
513                    del new_covobs[name]
514            cnames_loc = list(new_covobs.keys())
515        else:
516            cnames_loc = cnames
517        for name in cnames_loc:
518            res[i].names.append(name)
519            res[i].shape[name] = 1
520            res[i].idl[name] = []
521        res[i]._covobs = new_covobs
522
523    if symbol:
524        for i in range(len(res)):
525            res[i].tag = symbol[i]
526            if res[i].tag == 'None':
527                res[i].tag = None
528    if not noempty:
529        _check(len(res[0].covobs.keys()) == nc)
530    if full_output:
531        retd = {}
532        tool = file_origin.get('tool', None)
533        if tool:
534            program = tool['name'] + ' ' + tool['version']
535        else:
536            program = ''
537        retd['program'] = program
538        retd['version'] = version
539        retd['who'] = file_origin['who']
540        retd['date'] = file_origin['date']
541        retd['host'] = file_origin['host']
542        retd['description'] = descriptiond
543        retd['enstags'] = enstags
544        retd['obsdata'] = res
545        return retd
546    else:
547        return res
548
549
550def read_dobs(fname, noempty=False, full_output=False, gz=True, separator_insertion=True):
551    """Import a list of Obs from an xml.gz file in the Zeuthen dobs format.
552
553    Tags are not written or recovered automatically.
554
555    Parameters
556    ----------
557    fname : str
558        Filename of the input file.
559    noemtpy : bool
560        If True, ensembles with no contribution to the Obs are not included.
561        If False, ensembles are included as written in the file.
562    full_output : bool
563        If True, a dict containing auxiliary information and the data is returned.
564        If False, only the data is returned as list.
565    gz : bool
566        If True, assumes that data is gzipped. If False, assumes XML file.
567    separatior_insertion: str, int or bool
568        str: replace all occurences of "separator_insertion" within the replica names
569        by "|%s" % (separator_insertion) when constructing the names of the replica.
570        int: Insert the separator "|" at the position given by separator_insertion.
571        True (default): separator "|" is inserted after len(ensname), assuming that the
572        ensemble name is a prefix to the replica name.
573        None or False: No separator is inserted.
574    """
575
576    if not fname.endswith('.xml') and not fname.endswith('.gz'):
577        fname += '.xml'
578    if gz:
579        if not fname.endswith('.gz'):
580            fname += '.gz'
581        with gzip.open(fname, 'r') as fin:
582            content = fin.read().decode('utf-8')
583    else:
584        if fname.endswith('.gz'):
585            warnings.warn("Trying to read from %s without unzipping!" % fname, UserWarning)
586        with open(fname, 'r', encoding='utf-8') as fin:
587            content = fin.read()
588
589    # open and read gzipped xml file
590    infile = gzip.open(fname)
591    content = infile.read()
592
593    return import_dobs_string(content, noempty, full_output, separator_insertion=separator_insertion)
594
595
596def _dobsdict_to_xmlstring(d):
597    if isinstance(d, dict):
598        iters = ''
599        for k in d:
600            if k.startswith('#value'):
601                for li in d[k]:
602                    iters += li
603                return iters + '\n'
604            elif k.startswith('#'):
605                for li in d[k]:
606                    iters += li
607                iters = '<array>\n' + iters + '<%sarray>\n' % ('/')
608                return iters
609            if isinstance(d[k], dict):
610                iters += '<%s>\n' % (k) + _dobsdict_to_xmlstring(d[k]) + '<%s%s>\n' % ('/', k)
611            elif isinstance(d[k], str):
612                if len(d[k]) > 100:
613                    iters += '<%s>\n ' % (k) + d[k] + ' \n<%s%s>\n' % ('/', k)
614                else:
615                    iters += '<%s> ' % (k) + d[k] + ' <%s%s>\n' % ('/', k)
616            elif isinstance(d[k], list):
617                tmps = ''
618                if k in ['edata', 'cdata']:
619                    for i in range(len(d[k])):
620                        tmps += '<%s>\n' % (k) + _dobsdict_to_xmlstring(d[k][i]) + '</%s>\n' % (k)
621                else:
622                    for i in range(len(d[k])):
623                        tmps += _dobsdict_to_xmlstring(d[k][i])
624                iters += tmps
625            elif isinstance(d[k], (int, float)):
626                iters += '<%s> ' % (k) + str(d[k]) + ' <%s%s>\n' % ('/', k)
627            elif not d[k]:
628                return '\n'
629            else:
630                raise Exception('Type', type(d[k]), 'not supported in export!')
631    else:
632        raise Exception('Type', type(d), 'not supported in export!')
633    return iters
634
635
636def _dobsdict_to_xmlstring_spaces(d, space='  '):
637    s = _dobsdict_to_xmlstring(d)
638    o = ''
639    c = 0
640    cm = False
641    for li in s.split('\n'):
642        if li.startswith('<%s' % ('/')):
643            c -= 1
644            cm = True
645        for i in range(c):
646            o += space
647        o += li + '\n'
648        if li.startswith('<') and not cm:
649            if not '<%s' % ('/') in li:
650                c += 1
651        cm = False
652    return o
653
654
655def create_dobs_string(obsl, name, spec='dobs v1.0', origin='', symbol=[], who=None, enstags={}):
656    """Generate the string for the export of a list of Obs or structures containing Obs
657    to a .xml.gz file according to the Zeuthen dobs format.
658
659    Tags are not written or recovered automatically. The separator |is removed from the replica names.
660
661    Parameters
662    ----------
663    obsl : list
664        List of Obs that will be exported.
665        The Obs inside a structure do not have to be defined on the same set of configurations,
666        but the storage requirement is increased, if this is not the case.
667    name : str
668        The name of the observable.
669    spec : str
670        Optional string that describes the contents of the file.
671    origin : str
672        Specify where the data has its origin.
673    symbol : list
674        A list of symbols that describe the observables to be written. May be empty.
675    who : str
676        Provide the name of the person that exports the data.
677    enstags : dict
678        Provide alternative enstag for ensembles in the form enstags = {ename: enstag}
679        Otherwise, the ensemble name is used.
680    """
681    od = {}
682    r_names = []
683    for o in obsl:
684        r_names += [name for name in o.names if name.split('|')[0] in o.mc_names]
685    r_names = sorted(set(r_names))
686    mc_names = sorted(set([n.split('|')[0] for n in r_names]))
687    for tmpname in mc_names:
688        if tmpname not in enstags:
689            enstags[tmpname] = tmpname
690    ne = len(set(mc_names))
691    cov_names = []
692    for o in obsl:
693        cov_names += list(o.cov_names)
694    cov_names = sorted(set(cov_names))
695    nc = len(set(cov_names))
696    od['OBSERVABLES'] = {}
697    od['OBSERVABLES']['SCHEMA'] = {'NAME': 'lattobs', 'VERSION': '1.0'}
698    if who is None:
699        who = getpass.getuser()
700    od['OBSERVABLES']['origin'] = {
701        'who': who,
702        'date': str(datetime.datetime.now())[:-7],
703        'host': socket.gethostname(),
704        'tool': {'name': 'pyerrors', 'version': pyerrorsversion.__version__}}
705    od['OBSERVABLES']['dobs'] = {}
706    pd = od['OBSERVABLES']['dobs']
707    pd['spec'] = spec
708    pd['origin'] = origin
709    pd['name'] = name
710    pd['array'] = {}
711    pd['array']['id'] = 'val'
712    pd['array']['layout'] = '1 f%d' % (len(obsl))
713    osymbol = ''
714    if symbol:
715        if not isinstance(symbol, list):
716            raise Exception('Symbol has to be a list!')
717        if not (len(symbol) == 0 or len(symbol) == len(obsl)):
718            raise Exception('Symbol has to be a list of lenght 0 or %d!' % (len(obsl)))
719        osymbol = symbol[0]
720        for s in symbol[1:]:
721            osymbol += ' %s' % s
722        pd['array']['symbol'] = osymbol
723
724    pd['array']['#values'] = ['  '.join(['%1.16e' % o.value for o in obsl])]
725    pd['ne'] = '%d' % (ne)
726    pd['nc'] = '%d' % (nc)
727    pd['edata'] = []
728    for name in mc_names:
729        ed = {}
730        ed['enstag'] = enstags[name]
731        onames = sorted([n for n in r_names if (n.startswith(name + '|') or n == name)])
732        nr = len(onames)
733        ed['nr'] = nr
734        ed[''] = []
735
736        for r in range(nr):
737            ad = {}
738            repname = onames[r]
739            ad['id'] = repname.replace('|', '')
740            idx = _merge_idx([o.idl.get(repname, []) for o in obsl])
741            Nconf = len(idx)
742            layout = '%d i f%d' % (Nconf, len(obsl))
743            ad['layout'] = layout
744            data = ''
745            counters = [0 for o in obsl]
746            for ci in idx:
747                data += '%d ' % ci
748                for oi in range(len(obsl)):
749                    o = obsl[oi]
750                    if repname in o.idl:
751                        if counters[oi] < 0:
752                            data += '0 '
753                            continue
754                        if o.idl[repname][counters[oi]] == ci:
755                            num = o.deltas[repname][counters[oi]]
756                            if num == 0:
757                                data += '0 '
758                            else:
759                                data += '%1.16e ' % (num)
760                            counters[oi] += 1
761                            if counters[oi] >= len(o.idl[repname]):
762                                counters[oi] = -1
763                        else:
764                            data += '0 '
765                    else:
766                        data += '0 '
767                data += '\n'
768            ad['#data'] = data
769            ed[''].append(ad)
770        pd['edata'].append(ed)
771
772        allcov = {}
773        for o in obsl:
774            for name in o.cov_names:
775                if name in allcov:
776                    if not np.array_equal(allcov[name], o.covobs[name].cov):
777                        raise Exception('Inconsistent covariance matrices for %s!' % (name))
778                else:
779                    allcov[name] = o.covobs[name].cov
780        pd['cdata'] = []
781        for name in cov_names:
782            cd = {}
783            cd['id'] = name
784
785            covd = {'id': 'cov'}
786            if allcov[name].shape == ():
787                ncov = 1
788                covd['layout'] = '1 1 f'
789                covd['#data'] = '%1.14e' % (allcov[name])
790            else:
791                shape = allcov[name].shape
792                assert (shape[0] == shape[1])
793                ncov = shape[0]
794                covd['layout'] = '%d %d f' % (ncov, ncov)
795                ds = ''
796                for i in range(ncov):
797                    for j in range(ncov):
798                        val = allcov[name][i][j]
799                        if val == 0:
800                            ds += '0 '
801                        else:
802                            ds += '%1.14e ' % (val)
803                    ds += '\n'
804                covd['#data'] = ds
805
806            gradd = {'id': 'grad'}
807            gradd['layout'] = '%d f%d' % (ncov, len(obsl))
808            ds = ''
809            for i in range(ncov):
810                for o in obsl:
811                    if name in o.covobs:
812                        val = o.covobs[name].grad[i]
813                        if val != 0:
814                            ds += '%1.14e ' % (val)
815                        else:
816                            ds += '0 '
817                    else:
818                        ds += '0 '
819            gradd['#data'] = ds
820            cd['array'] = [covd, gradd]
821            pd['cdata'].append(cd)
822
823    rs = '<?xml version="1.0" encoding="utf-8"?>\n' + _dobsdict_to_xmlstring_spaces(od)
824
825    return rs
826
827
828def write_dobs(obsl, fname, name, spec='dobs v1.0', origin='', symbol=[], who=None, enstags={}, gz=True):
829    """Export a list of Obs or structures containing Obs to a .xml.gz file
830    according to the Zeuthen dobs format.
831
832    Tags are not written or recovered automatically. The separator | is removed from the replica names.
833
834    Parameters
835    ----------
836    obsl : list
837        List of Obs that will be exported.
838        The Obs inside a structure do not have to be defined on the same set of configurations,
839        but the storage requirement is increased, if this is not the case.
840    fname : str
841        Filename of the output file.
842    name : str
843        The name of the observable.
844    spec : str
845        Optional string that describes the contents of the file.
846    origin : str
847        Specify where the data has its origin.
848    symbol : list
849        A list of symbols that describe the observables to be written. May be empty.
850    who : str
851        Provide the name of the person that exports the data.
852    enstags : dict
853        Provide alternative enstag for ensembles in the form enstags = {ename: enstag}
854        Otherwise, the ensemble name is used.
855    gz : bool
856        If True, the output is a gzipped XML. If False, the output is a XML file.
857    """
858
859    dobsstring = create_dobs_string(obsl, name, spec, origin, symbol, who, enstags=enstags)
860
861    if not fname.endswith('.xml') and not fname.endswith('.gz'):
862        fname += '.xml'
863
864    if gz:
865        if not fname.endswith('.gz'):
866            fname += '.gz'
867
868        fp = gzip.open(fname, 'wb')
869        fp.write(dobsstring.encode('utf-8'))
870    else:
871        fp = open(fname, 'w', encoding='utf-8')
872        fp.write(dobsstring)
873    fp.close()