brightway-lca / brightway2-calc Goto Github PK

Original report by Caroline Nadel (Bitbucket: cazcazn, ).

Following the tutorials. Everytime I try to do a Monte Carlo as documented, an exception ValueError: left > mode is thrown. This happens with both tutorials 2 and 3 so far.

Stack trace:

ValueError Traceback (most recent call last)
in ()
2 mc.load_data()
3 for x in range(10):
----> 4 print mc.next()

.../env/lib/python2.7/site-packages/bw2calc/monte_carlo.pyc in next(self)
64 if not hasattr(self, "tech_rng"):
65 raise NameError("Must run load_data before making calculations")
---> 66 self.rebuild_technosphere_matrix(self.tech_rng.next())
67 self.rebuild_biosphere_matrix(self.bio_rng.next())
68 if self.lcia:

.../env/lib/python2.7/site-packages/stats_arrays/random.pyc in next(self)
181 1,
182 self.random,
--> 183 self.maximum_iterations
184 )
185 if len(random_data.shape) == 2:

.../env/lib/python2.7/site-packages/stats_arrays/distributions/base.pyc in bounded_random_variables(cls, params, size, seeded_random, maximum_iterations)
314 maximum_iterations=None):
315 """No bounds checking because the bounds do not exclude any of the distribution."""
--> 316 return cls.random_variables(params, size, seeded_random)
317
318 @classmethod

.../env/lib/python2.7/site-packages/stats_arrays/distributions/geometric.pyc in random_variables(cls, params, size, seeded_random)
68 params['loc'], # Mode
69 params['maximum'], # Right
---> 70 size=(size, params.shape[0])).T
71
72 @classmethod

mtrand.pyx in mtrand.RandomState.triangular (numpy/random/mtrand/mtrand.c:19950)()

ValueError: left > mode

Current logging is rarely useful and doesn't meet modern needs. Please vote on the core features a reworked logging should include:

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

Including:

• activity_dict
• product_dict
• reverse_dict

When I try to use MultiLCA, I get the error: PyPardisoError: The Pardiso solver failed with error code -3. See Pardiso documentation for details.

I tried to update brightway2, ran into the issue here (AttributeError: module 'bw2calc' has no attribute 'ComparativeMonteCarlo'). I tried to follow the answer tngTUDIR so I updated python in the process to its 3.10 version. And now my calculations that were working before trying to update brightway2 leads to the error described above.

It would be sensible for multiple LCIA methods to be calculated at once - however, it is unclear exactly what this would look like. What does it look like to the end user? For example, is LCA.method now LCA.methods? Is LCA.characterization_matrix a list? What about LCA.score(s)?

One option would be for all method attributes to become lists, but behave like a single element if no explicit index is provided. However, this might be a bit too much magic. On the other hand, a switch from LCA.method to LCA.methods (which would always be a list) will break all existing Brightway code. So maybe the LCA class should always only support one impact category, and a new class should be added for multiple methods and functional units?

I don't know. Please provide your opinions, and, more helpfully, some actual interface ideas!

Currently, bw-calc only calculates the solution to the inventory problem, i.e. s in As = f. This is sufficient to get LCA results and do contribution analysis. But it is not sufficient to generate Sankey diagrams that show how impacts build up across the life cycle. The work-around that has been used, e.g. in the Activity Browser, is to use the brightway graph-traversal. This works, but it is not ideal for generating the Sankey diagrams as one graph traversal has to be done for each reference flow and impact category. Having the matrix inverse (A-1) would cost some up-front calculation time, but then make Sankey diagram calculations orders of magnitude quicker (e.g. like in SimaPro). To my knowledge, one of the reasons why calculations in OpenLCA and SimaPro take longer is because the matrix inverses are calculated. OpenLCA has found an elegant solution, which could be copied in bw. You can chose how you want to calculate the LCA results (there is a "quick" way (without matrix inverse, not giving you Sankeys) and an "analysis" way (calculating the inverse and providing Sankey results)).

So what I am practically suggesting is to include a method (e.g. as in bw.LCA(... ,... , calculate_matrix_inverse=False)) that includes a matrix inversion. Default should be "False" to get quick LCA results, but the option "True" should do it (or at least provide) the inverted A matrix.

By the way, this can have other benefits as well:

• Sankey diagrams
• once you have the matrix inverse you can calculate LCA results for all processes in your database MUCH quicker

Sorry, I had limited time to write this up, but it would be great if bw could also return the inverse technology matrix (and possibly also do things with it)

This is an eternal issue in Brightway development, one can find artifacts of previous report formats scattered around.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

Logging framework gist here.

Need new input argument for all LCA classes: logpath. This can be either a filepath or a directory (in which case, a log filename is generated). We then need to log the following:

• Filepaths of all processed arrays for each matrix
• Hash of each processed array. This is fast, ~one millionth of a second per array.
• Functional unit
• Seed value for RNGs if Monte Carlo

Brightway2-data will be updated to include a new datastore for these logfiles, and utility functions to read the logs and construct the necessary arguments to reproduce the calculations.

Original report by Pascal Lesage (Bitbucket: MPa, ).

build_demand_array fails:

#!python

db = bw.Database('ecoinvent 3_2 CutOff')
act = db.random()
myMethod = bw.methods.random()
myLCA = bw.LCA({act:1}, myMethod)
demand_array = myLCA.build_demand_array()

AttributeError: 'LCA' object has no attribute 'product_dict'

Specifying the product does not help:

#!python

demand_array = myLCA.build_demand_array(demand = {act.get('reference product')})

System: Windows 10. BW updated 2016/04/15

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

See recent commit to scikit-umfpack: scikit-umfpack/scikit-umfpack@3d5cd82

Original report by Miguel F.Ast (Bitbucket: Miguel_fast, ).

For some reason MultimonteCarlo method appears to provide identical solutions for all the iterations on the same activity e.g.

import brightway2 as bw
act_dict_list=[{bw.Database('ecoinvent34_consequential').random():1},
           {bw.Database('ecoinvent34_consequential').random():1},
           {bw.Database('ecoinvent34_consequential').random():1}]

ipcc2013=[m for m in bw.methods if 'IPCC' in m[0]
                    and ('2013') in str(m)
                    and 'GWP 100' in str(m)
                    and 'no LT' not in str(m)][0]

mmc=bw.MultiMonteCarlo(act_dict_list,method=ipcc2013,iterations=10)
r=mmc.calculate()

Good morning,

I have been trying to install brightway2 in the framework of my Phd thesis but I have a problem when importing it.

I attach all the steps I followed for the installation.

=> I installed Miniconda3

=> I create a new environment with python 3.7

=> Information from the newly-created environment…

=> conda install -y -q -c conda-forge -c cmutel -c haasad brightway2 jupyter

=> conda install -y -q pywin32

=> conda clean –tipsy

Here I get some warnings… I do not know if they are important or not.

=> conda update -c cmutel -c conda-forge bw2data bw2calc bw2io

=> pip install numpy
=> conda install pandas

I listed all the packages that I have in my environment. I was thinking that maybe I was not using an updated version of some packages. I dug into that but I could not find anything.

In my opinion, the creation of the environment and the importation of all the packages were fine.
However, when I try to import brightway2 in a jupyter notebook, I get this error…

I'm sorry to bother you with this issue but I am beginner with brightway and I do not have any clue about how I could solve the problem. I tried many times to create new environments, update packages etc... but it still doesn't work.

Do you have any clue about it?

Thanks a lot in advance for your time

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

There is not easy way to fix this for the time being, however. Need to maintain presample behaviour where the last value is chosen (instead of adding as in matrix construction).

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

When installing the latest version of the brightway2 metapackage (2.4.1) or also just the latest stable release of bw2calc (1.8.0), I get the following error: AttributeError: module 'bw2calc' has no attribute 'ComparativeMonteCarlo'

To reproduce:

• fresh install

conda create -n ab705_4 -c conda-forge -c cmutel python=3.9 brightway2
conda activate ab705_4

• only importing bw2calc works fine:

python -c "import bw2calc; print(bw2calc.__version__)" 
(1, 8, 0)

• importing brightway2 fails:

python -c "import brightway2 as bw"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/adrian/miniconda3/envs/ab705_4/lib/python3.9/site-packages/brightway2/__init__.py", line 3, in <module>
    from bw2calc import *
AttributeError: module 'bw2calc' has no attribute 'ComparativeMonteCarlo'

• also:

python -c "from bw2calc import *"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
AttributeError: module 'bw2calc' has no attribute 'ComparativeMonteCarlo'

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

If you have a database 'foo', and accidentally set up an LCA with LCA({Database("bar").random(): 1}), you get the following error:

#!python

>>> mc = MonteCarloLCA({Database("ecoinvent 3.2 cutoff").random(): 1}, methods.random())
Warning: This database is empty
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/jupyter/libs/python-3.5.1/lib/python3.5/site-packages/bw2calc/monte_carlo.py", line 21, in __init__
    **kwargs)
  File "/home/jupyter/libs/python-3.5.1/lib/python3.5/site-packages/bw2calc/lca.py", line 66, in __init__
    self.get_array_filepaths()
  File "/home/jupyter/libs/python-3.5.1/lib/python3.5/site-packages/bw2calc/lca.py", line 71, in get_array_filepaths
    get_database_filepaths(self.demand, self._databases),
  File "/home/jupyter/libs/python-3.5.1/lib/python3.5/site-packages/bw2calc/utils.py", line 64, in get_database_filepaths
    dbs = set.union(*[Database(key[0]).find_graph_dependents() for key in functional_unit])
  File "/home/jupyter/libs/python-3.5.1/lib/python3.5/site-packages/bw2calc/utils.py", line 64, in <listcomp>
    dbs = set.union(*[Database(key[0]).find_graph_dependents() for key in functional_unit])
TypeError: 'NoneType' object is not subscriptable

This does nothing to help you understand what the problem actually was.

I cannot calculate any lcia results with the latest version of brightway on my 64-bit Windows machine. I receive an error claiming that the row indices passed during matrix building contain negative values. Stepping through the code, I found that there are not really any negative numbers. The problem is caused by faulty integer interpretations. The array of row indices contains the number 4294967295 which is interpreted as -1 when cast to int32 (which scipy.sparse.coo_matrix seems to do). The number is introduced by function index_with_arrays in bw2calc/indexing.py. It contains the following lines:

index_array = np.zeros(keys.max() + 1) - 1
index_array[keys] = values

mask = array_from <= keys.max()
array_to[:] = -1 # array_to contains only 4294967295's
array_to[mask] = index_array[array_from[mask]]
array_to[array_to == -1] = MAX_SIGNED_32BIT_INT # replacement fails

Because the last line, the replacement step, fails the array still contains many 4294967295's when returned. For some weird reason, 4294967295 is not interpreted as -1 during the replacement but later during matrix building. Thus the error.

Possible solution (no error on my machine but didn't test it thoroughly yet):

index_array = np.zeros(keys.max() + 1) - 1
index_array[keys] = values
index_array[index_array == -1] = MAX_SIGNED_32BIT_INT

mask = array_from <= keys.max()
array_to[:] = MAX_SIGNED_32BIT_INT 
array_to[mask] = index_array[array_from[mask]]

I installed Brightway 2 in a new computer.

I tried to create a new project:
project.set_current (new)

project.set_current (new)

However, I can not create it. Please, look at the message below:

Thank you in advance.

`---------------------------------------------------------------------------``
IntegrityError Traceback (most recent call last)
~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in execute_sql(self, sql, params, commit)
3159 try:
-> 3160 cursor.execute(sql, params or ())
3161 except Exception:

IntegrityError: NOT NULL constraint failed: projectdataset.full_hash

During handling of the above exception, another exception occurred:

IntegrityError Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_16580/2692245280.py in
----> 1 projects.set_current("new")

~\Miniconda3\envs\BW2\lib\site-packages\bw2data\project.py in set_current(self, name, writable, update)
141 # for new metadata stores
142 self.read_only = False
--> 143 self.create_project(name)
144 self._reset_meta()
145 self._reset_sqlite3_databases()

~\Miniconda3\envs\BW2\lib\site-packages\bw2data\project.py in create_project(self, name, **kwargs)
210 if not ProjectDataset.select().where(
211 ProjectDataset.name == name).count():
--> 212 ProjectDataset.create(
213 data=kwargs,
214 name=name

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in create(cls, **query)
6391 def create(cls, **query):
6392 inst = cls(**query)
-> 6393 inst.save(force_insert=True)
6394 return inst
6395

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in save(self, force_insert, only)
6601 rows = self.update(**field_dict).where(self._pk_expr()).execute()
6602 elif pk_field is not None:
-> 6603 pk = self.insert(**field_dict).execute()
6604 if pk is not None and (self._meta.auto_increment or
6605 pk_value is None):

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in inner(self, database, *args, **kwargs)
1909 raise InterfaceError('Query must be bound to a database in order '
1910 'to call "%s".' % method.name)
-> 1911 return method(self, database, *args, **kwargs)
1912 return inner
1913

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in execute(self, database)
1980 @database_required
1981 def execute(self, database):
-> 1982 return self._execute(database)
1983
1984 def _execute(self, database):

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in _execute(self, database)
2759 self._returning = (self.table._primary_key,)
2760 try:
-> 2761 return super(Insert, self)._execute(database)
2762 except self.DefaultValuesException:
2763 pass

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in _execute(self, database)
2477 cursor = self.execute_returning(database)
2478 else:
-> 2479 cursor = database.execute(self)
2480 return self.handle_result(database, cursor)
2481

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in execute(self, query, commit, **context_options)
3171 ctx = self.get_sql_context(**context_options)
3172 sql, params = ctx.sql(query).query()
-> 3173 return self.execute_sql(sql, params, commit=commit)
3174
3175 def get_context_options(self):

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in execute_sql(self, sql, params, commit)
3165 else:
3166 if commit and not self.in_transaction():
-> 3167 self.commit()
3168 return cursor
3169

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in exit(self, exc_type, exc_value, traceback)
2931 new_type = self.exceptions[exc_type.name]
2932 exc_args = exc_value.args
-> 2933 reraise(new_type, new_type(exc_value, *exc_args), traceback)
2934
2935

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in reraise(tp, value, tb)
189 def reraise(tp, value, tb=None):
190 if value.traceback is not tb:
--> 191 raise value.with_traceback(tb)
192 raise value
193

~\Miniconda3\envs\BW2\lib\site-packages\peewee.py in execute_sql(self, sql, params, commit)
3158 cursor = self.cursor(commit)
3159 try:
-> 3160 cursor.execute(sql, params or ())
3161 except Exception:
3162 if self.autorollback and not self.in_transaction():

IntegrityError: NOT NULL constraint failed: projectdataset.full_hash`

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

The Pardiso solver seems to offer significant speedups compared to UMFPACK, at least in initial tests. A wrapper to use this library should be built using ctypes and the documentation.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

The method_filepath doesn't change, even when the method does.

Depends on #51 and #50. Supports #51.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

See CI pipeline.

Not yet replicated locally.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

If a sample is never defined, but only pre-computed vectors are passed in, then comparing to self.sample will fail.

So, have to choose between:

1. Drawing a sample when self.rng is constructed on line 46, or
2. Comparing to length of array used to populate RNG, or
3. Doing a different comparison completely, i.e. 1-dimensional and fits into the self.positions.

Not sure which approach is best.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

Like, very out of date...

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

Needed for remote calculations. For example, ParallelMonteCarlo should only be passing processed array file paths instead of project, etc.

Multiple examples would be nice, e.g. a lambda function, a Docker container, or a native server install. Probably needs a library with a defined API (presumably REST).

Probably depends on #50.

Original report by Pascal Lesage (Bitbucket: MPa, ).

Despite apparently accepting demand=None, LCA.redo_lci() and LCA.redo_lcia() will fail if no demand argument is provided.

>>> import brightway2 as bw
>>> lca = bw.LCA({('some', 'act'):1})
>>> lca.lci()
>>> lca.redo_lci()

Traceback (most recent call last):
  File "C:\mypy\anaconda\envs\bw2020\lib\site-packages\IPython\core\interactiveshell.py", line 3331, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-25-82c6d21a2d04>", line 1, in <module>
    lca.redo_lci()
  File "C:\mypy\anaconda\envs\bw2020\lib\site-packages\bw2calc\lca.py", line 516, in redo_lci
    self.logger.info("Redoing LCI", extra={'demand': wrap_functional_unit(demand or self.demand)})
  File "C:\mypy\anaconda\envs\bw2020\lib\site-packages\bw2calc\utils.py", line 219, in wrap_functional_unit
    for key, amount in dct.items():
AttributeError: 'NoneType' object has no attribute 'items'

The problem is here (): and here. The line self.demand = demand should be in the if block.

I will provide a PR if there is no objection.

Original report by Pedro Anchieta (Bitbucket: phanchieta, GitHub: phanchieta).

Version 1.7.4:

‌

def load_arrays(paths):
    """Load the numpy arrays in list of filepaths ``paths``."""
    assert all(os.path.isfile(fp) for fp in paths)
    return np.hstack([np.load(path) for path in sorted(paths)])

was changed (in version 1.7.5) to

‌

...
for obj in objs:
    if isinstance(obj, np.ndarray):
        # we're done here as the object is already a numpy array
        arrays.append(obj.copy())
    else:
        # treat object as loadable by numpy and try to load it from disk
        arrays.append(np.load(obj))
return np.hstack(arrays)

List "arrays" appends unsorted objects and causes errors when using file paths.

I have a problem when for an example I try to run a montecarlo results for an activity of a database with a single activity. It throws a value error ValueError: A and x have incompatible dimensions after the first iteration. Here is the code to reproduce it. The problem does not happen if the database has another activity.

pm25_key=('biosphere3', '051aaf7a-6c1a-4e86-999f-85d5f0830df6')

act1_key=('test_1_act','activity_1')

biosphere_exchange_1={'amount':1,
                    'input':pm25_key,
                    'output':act1_key,
                    'type':'biosphere',
                    'uncertainty type': 0}

production_exchange_1={'amount':1,
                     'input':act1_key,
                     'output':act1_key,
                     'type':'production',
                     'uncertainty type':0}

act_1_dict={'name':'activity_1',
             'unit':'megajoule',
             'exchanges':[production_exchange_1,biosphere_exchange_1]}

database_dict={act1_key:act_1_dict}

db=bw.Database('test_1_act')

db.write(database_dict)

a1=bw.get_activity(act1_key)

# montecarlo, problem after first iteration
mc1a=bw.MonteCarloLCA({a1:1},bw.methods.random())

next(mc1a)

next(mc1a)

Original report by Pascal Lesage (Bitbucket: MPa, ).

When instantiating an LCA, the LCA-level seed is passed to the matrix_presamples:

#!python

self.presamples = [MatrixPresamples(presamples, self.seed)]    

This overwrites presamples-level seeds.
It would be more useful for presamples seeds to be independent of LCA seeds.

Original report by Pascal Lesage (Bitbucket: MPa, ).

Right now, seeds are passed when matrix_presamples are loaded. They should be passed when presamples are created. This is the correct time to distinguish between randomly sampled presamples and sequentially sampled presamples. It also improves reproducibility.

Help - I don't know where to ask this question.

I think BW2 could benefit from being able to calculate results outputting a function instead of a number.

eg. GWP(x) = a x + b

In my mind this should be doable, and I would like to contribute to BW2 by developing this feature. However I don't really know where to start. Any pointers would be appreciated.

From what I can see, Activity Browser does not allow parameters to have unknown values

Hello, I have recently re-installed brightway and I get an error when I trying to import bw2calc (v1.8.0) on a windows machine. It seems to be due to library used by the solver Pardiso (v 0.2.2 installed) ... I am not sure how to address this.

 FileNotFoundError                         Traceback (most recent call last)
<ipython-input-1-a3ddbb8a34e8> in <module>
----> 1 import brightway2 as bw

~\Anaconda3\envs\bw\lib\site-packages\brightway2\__init__.py in <module>
      1 # -*- coding: utf-8 -*
      2 from bw2data import *
----> 3 from bw2calc import *
      4 from bw2io import *
      5

~\Anaconda3\envs\bw\lib\site-packages\bw2calc\__init__.py in <module>
     24 __version__ = (1, 8, 0)
     25
---> 26 from .lca import LCA
     27 from .dense_lca import DenseLCA
     28 from .independent_lca import IndependentLCAMixin

~\Anaconda3\envs\bw\lib\site-packages\bw2calc\lca.py in <module>
     27
     28 try:
---> 29     from pypardiso import factorized, spsolve
     30 except ImportError:
     31     from scipy.sparse.linalg import factorized, spsolve

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _find_and_load(name, import_)

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _find_and_load_unlocked(name, import_)

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _load_unlocked(spec)

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _load_backward_compatible(spec)

<frozen zipimport> in load_module(self, fullname)

~\Anaconda3\envs\bw\lib\site-packages\pypardiso-0.2.2-py3.6.egg\pypardiso\__init__.py in <module>
      2
      3 from .pardiso_wrapper import PyPardisoSolver
----> 4 from .scipy_aliases import spsolve, factorized
      5 from .scipy_aliases import pypardiso_solver as ps
      6

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _find_and_load(name, import_)

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _find_and_load_unlocked(name, import_)

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _load_unlocked(spec)

~\Anaconda3\envs\bw\lib\importlib\_bootstrap.py in _load_backward_compatible(spec)

<frozen zipimport> in load_module(self, fullname)

~\Anaconda3\envs\bw\lib\site-packages\pypardiso-0.2.2-py3.6.egg\pypardiso\scipy_aliases.py in <module>
      7 # pypardsio_solver is used for the 'spsolve' and 'factorized' functions. Python crashes on windows if multiple
      8 # instances of PyPardisoSolver make calls to the Pardiso library
----> 9 pypardiso_solver = PyPardisoSolver()
     10
     11

~\Anaconda3\envs\bw\lib\site-packages\pypardiso-0.2.2-py3.6.egg\pypardiso\pardiso_wrapper.py in __init__(self, mtype, phase, size_limit_storage)
     64             self.libmkl = ctypes.CDLL('libmkl_rt.dylib')
     65         elif sys.platform == 'win32':
---> 66             self.libmkl = ctypes.CDLL('mkl_rt.dll')
     67         else:
     68             self.libmkl = ctypes.CDLL('libmkl_rt.so')

~\Anaconda3\envs\bw\lib\ctypes\__init__.py in __init__(self, name, mode, handle, use_errno, use_last_error, winmode)
    379
    380         if handle is None:
--> 381             self._handle = _dlopen(self._name, mode)
    382         else:
    383             self._handle = handle

FileNotFoundError: Could not find module 'mkl_rt.dll' (or one of its dependencies). Try using the full path with constructor syntax.

Hi

I was having some trouble with to_dataframe and I see now it is commented. In case this method will be used at any point, I think stacked.sort() should be replaced by something like stacked = stacked[:,stacked[0,:].argsort()] (or anything more elegant).. otherwise the ranking is not well sorted.

The URLs in the README point to former domain brightwaylca.org.

Original report by Pascal Lesage (Bitbucket: MPa, ).

As reported here, Brightway and stats_arrays uncertainty dictionaries are different: one uses the key uncertainty_type, and the other uncertainty type.

This sometimes causes confusion.

It also sometimes appear to cause errors when defining new distributions (code available upon request).

Harmonization would be great.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

There are significant differences between the behaviour of multiprocessing on Unix and Windows. This has caused issues with Monte Carlo calculations.

Need to set the project in each Monte Caro worker process.

Original report by Pascal Lesage (Bitbucket: MPa, ).

Said attributes would be informed by presamples. Useful for sensitivity analysis.

Hello everyone,
Recently I am struggling with a problem of brightway import. I am performing an LCA project for my master thesis and it is stuck. Since yesterday, python/jupyternotebook is giving the following problem. I need to find a solution asap. Is there anyone who can help me? Any advice or answer would be awesome helpful.

Thank u!

### Input
import brightway2 as bw
import os
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt

### Output

ImportError Traceback (most recent call last)
File ~\anaconda3\envs\MG\lib\site-packages\bw2data\serialization.py:242, in PickledDict.deserialize(self)
241 try:
--> 242 return self.unpack(pickle.load(open(self.filepath, "rb")))
243 except ImportError:

File ~\anaconda3\envs\MG\lib\site-packages\bw2data\backends_init_.py:5, in
3 from eight import *
----> 5 from .base import LCIBackend
6 from .peewee import SQLiteBackend

File ~\anaconda3\envs\MG\lib\site-packages\bw2data\backends\base.py:5, in
3 from eight import *
----> 5 from .. import (
6 config,
7 databases,
8 geomapping,
9 mapping,
10 projects,
11 )
12 from ..data_store import ProcessedDataStore

**ImportError: cannot import name 'databases' from partially initialized module 'bw2data' (most likely due to a circular import) (C:\Users\cansu\anaconda3\envs\MG\lib\site-packages\bw2data_init_.py)

During handling of the above exception, another exception occurred:**

PickleError Traceback (most recent call last)
Input In [1], in <cell line: 1>()
----> 1 import brightway2 as bw
2 import os # to use "operating system dependent functionality"
3 import numpy as np # "the fundamental package for scientific computing with Python"

File ~\anaconda3\envs\MG\lib\site-packages\brightway2_init_.py:2, in
1 # -- coding: utf-8 -
----> 2 from bw2data import *
3 from bw2calc import *
4 from bw2io import *

File ~\anaconda3\envs\MG\lib\site-packages\bw2data_init_.py:35, in
33 from .project import projects
34 from .utils import set_data_dir
---> 35 from .meta import (
36 dynamic_calculation_setups,
37 calculation_setups,
38 databases,
39 geomapping,
40 mapping,
41 methods,
42 normalizations,
43 preferences,
44 weightings,
45 )
47 # Add metadata class instances to global list of serialized metadata
48 config.metadata.extend([
49 dynamic_calculation_setups,
50 calculation_setups,
(...)
57 weightings,
58 ])

File ~\anaconda3\envs\MG\lib\site-packages\bw2data\meta.py:192, in
190 preferences = Preferences()
191 weightings = WeightingMeta()
--> 192 calculation_setups = CalculationSetups()
193 dynamic_calculation_setups = DynamicCalculationSetups()

File ~\anaconda3\envs\MG\lib\site-packages\bw2data\serialization.py:123, in SerializedDict.init(self, dirpath)
118 raise NotImplemented("SerializedDict must be subclassed, and the filename must be set.")
119 self.filepath = os.path.join(
120 dirpath or projects.dir,
121 self.filename
122 )
--> 123 self.load()

File ~\anaconda3\envs\MG\lib\site-packages\bw2data\serialization.py:128, in SerializedDict.load(self)
126 """Load the serialized data. Creates the file if not yet present."""
127 try:
--> 128 self.data = self.deserialize()
129 except IOError:
130 # Create if not present
131 self.data = {}

File ~\anaconda3\envs\MG\lib\site-packages\bw2data\serialization.py:245, in PickledDict.deserialize(self)
243 except ImportError:
244 TEXT = "Pickle deserialization error in file '%s'" % self.filepath
--> 245 raise PickleError(TEXT)

PickleError: Pickle deserialization error in file 'C:\Users\cansu\AppData\Local\pylca\Brightway3\default.c21f969b5f03d33d43e04f8f136e7682\setups.pickle'

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

This will be quite some work:

1. We need to be able to package up the processed files for a given functional unit and method (and weighting, normalization). This should be a single archive file.

2. Translate functional unit, etc. to the correct filenames and indices.

3. Write bw2remote, which is a flask application that can accept the files and a JSON payload.

4. bw2remote should then stream Monte Carlo results back to a notebook widget that will produce a dynamic D3 histogram. See: https://github.com/mbostock/d3/wiki/Histogram-Layout, http://stackoverflow.com/questions/22052694/how-to-update-d3-js-bar-chart-with-new-data, http://blog.thedataincubator.com/2015/08/embedding-d3-in-an-ipython-notebook/

bw2remote should use rq and redis: https://redis-py.readthedocs.org/en/latest/, http://python-rq.org/docs/, and a Queue during multiprocessing.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

If the technosphere matrix is changed, the factorized solver won't know about it. See discussion and (hopefully eventual solution) here on pypardiso issue.

It is currently not possible to switch methods using lca.switch_method. The reason: lca.switch_method loads packages and calls lca.load_lcia_data(packages) successfully. However, when calling lca.lcia() thereafter, lca.load_lcia_data() is again called - this time without arguments. This time self.packages is loaded, which still holds the old method data because it was never overwritten during switch_method.

Current implementation:

    def switch_method(self, method):
        """Switch to LCIA method `method`"""
        try:
            _, data_objs, _ = prepare_lca_inputs(method=method)
            packages = [load_package(obj) for obj in data_objs]
        except AssertionError:
            packages = method
        self.method = method
        self.load_lcia_data(packages)


    def lcia(self):
        assert hasattr(self, "inventory"), "Must do lci first"
        # assert self.method, "Must specify a method to perform LCIA"
        if not self.dicts.biosphere:
            raise EmptyBiosphere

        self.load_lcia_data()
        self.lcia_calculation()

Proposed fix:

```python

    def switch_method(self, method):
        """Switch to LCIA method `method`"""
        try:
            _, data_objs, _ = prepare_lca_inputs(method=method)
            packages = [load_package(obj) for obj in data_objs]
        except AssertionError:
            packages = method
        self.method = method
        self.packages = packages
        self.load_lcia_data(packages)


    def lcia(self):
        assert hasattr(self, "inventory"), "Must do lci first"
        # assert self.method, "Must specify a method to perform LCIA"
        if not self.dicts.biosphere:
            raise EmptyBiosphere

        self.load_lcia_data()
        self.lcia_calculation()

Additionally, one could introduce a flag to prevent load_lcia_data from being called twice.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

Currently this is impossible, because only matrix presamples are included when loading presamples. Seems like the most reasonable approach would be to override __init__, and append the necessary space onto self.sample. Note that the current implementation doesn't preserve self.sample; it could be that a wrapper class would be easier in this case.

Should be compatible with existing MultiLCA code as far as possible.

Original report by Chris Mutel (Bitbucket: cmutel, GitHub: cmutel).

UMFPACK is significantly faster than the default solver, but is a huge pain to package on Windows.

Packaging for 2.7 requires an ancient Microsoft compiler, and is out of scope for this ticket.

Current progress: Using Suitesparse-METIS for Windows, I can compile SuiteSparse, but don't understand how to "install" it so that the paths are correct for scikit-umfpack/numpy to find.

Also still not sure if the eventual Windows will include the SuiteSparse DLLs...

Current failure on Appveyor.

Branch of scikit-umfpack on Github.

See also (in no particular order):

• https://packaging.python.org/en/latest/appveyor/
• http://timcera.bitbucket.org/posts/appveyor_finally.html
• https://github.com/ogrisel/python-appveyor-demo
• http://stackoverflow.com/questions/24071491/how-can-i-make-a-python-wheel-from-an-existing-native-library
• https://bitbucket.org/pygame/pygame/pull-requests/45/create-python-wheel-builds-using-appveyor/diff
• https://github.com/numpy/windows-wheel-builder
• https://github.com/numpy/numpy/wiki/Whats-with-Windows-builds

Could also think about building wheels for Linux and OS X automatically, but one thing at a time. See travis config, just need to build wheels.