Source code for flopy.utils.binaryfile

"""
Module to read MODFLOW binary output files.  The module contains four
important classes that can be accessed by the user.

*  HeadFile (Binary head file.  Can also be used for drawdown)
*  HeadUFile (Binary MODFLOW-USG unstructured head file)
*  UcnFile (Binary concentration file from MT3DMS)
*  CellBudgetFile (Binary cell-by-cell flow file)

"""

import tempfile
import warnings
from os import PathLike
from pathlib import Path
from shutil import move
from typing import Optional, Union

import numpy as np
import pandas as pd

from flopy.discretization.modeltime import ModelTime

from ..datafile import Header, LayerFile
from ..gridutil import get_lni


[docs]class BinaryHeader(Header): """ Represents data headers for binary output files. Parameters ---------- bintype : str, default None Type of file being opened. Accepted values are 'head' and 'ucn'. precision : str, default 'single' Precision of floating point data in the file. """ def __init__(self, bintype=None, precision="single"): super().__init__(bintype, precision)
[docs] def set_values(self, **kwargs): """ Set values using kwargs """ ikey = [ "ntrans", "kstp", "kper", "ncol", "nrow", "ilay", "ncpl", "nodes", "m1", "m2", "m3", ] fkey = ["pertim", "totim"] ckey = ["text"] for k in ikey: if k in kwargs.keys(): try: self.header[0][k] = int(kwargs[k]) except: print(f"{k} key not available in {self.header_type} header dtype") for k in fkey: if k in kwargs.keys(): try: self.header[0][k] = float(kwargs[k]) except: print(f"{k} key not available in {self.header_type} header dtype") for k in ckey: if k in kwargs.keys(): # Convert to upper case to be consistent case used by MODFLOW # text strings. Necessary to work with HeadFile and UcnFile # routines ttext = kwargs[k].upper() # trim a long string if len(ttext) > 16: text = ttext[0:16] # pad a short string elif len(ttext) < 16: text = f"{ttext:<16}" # the string is just right else: text = ttext self.header[0][k] = text else: self.header[0][k] = "DUMMY TEXT"
[docs] @staticmethod def set_dtype(bintype=None, precision="single"): """ Set the dtype """ header = Header(filetype=bintype, precision=precision) return header.dtype
[docs] @staticmethod def create(bintype=None, precision="single", **kwargs): """ Create a binary header """ header = BinaryHeader(bintype=bintype, precision=precision) if header.get_dtype() is not None: header.set_values(**kwargs) return header.get_values()
[docs]def binaryread_struct(file, vartype, shape=(1,), charlen=16): """ Read text, a scalar value, or an array of values from a binary file. file : file object is an open file object vartype : type is the return variable type: str, numpy.int32, numpy.float32, or numpy.float64 shape : tuple is the shape of the returned array (shape(1, ) returns a single value) for example, shape = (nlay, nrow, ncol) charlen : int is the length of the text string. Note that string arrays cannot be returned, only multi-character strings. Shape has no affect on strings. .. deprecated:: 3.8.0 Use :meth:`binaryread` instead. """ import struct warnings.warn( "binaryread_struct() is deprecated; use binaryread() instead.", DeprecationWarning, ) # store the mapping from type to struct format (fmt) typefmtd = {np.int32: "i", np.float32: "f", np.float64: "d"} # read a string variable of length charlen if vartype == str: result = file.read(charlen * 1) # read other variable types else: fmt = typefmtd[vartype] # find the number of bytes for one value numbytes = vartype(1).nbytes # find the number of values nval = np.prod(shape) fmt = str(nval) + fmt s = file.read(numbytes * nval) result = struct.unpack(fmt, s) if nval == 1: result = vartype(result[0]) else: result = np.array(result, dtype=vartype) result = np.reshape(result, shape) return result
[docs]def binaryread(file, vartype, shape=(1,), charlen=16): """ Read character bytes, scalar or array values from a binary file. Parameters ---------- file : file object is an open file object vartype : type is the return variable type: bytes, numpy.int32, numpy.float32, or numpy.float64. Using str is deprecated since bytes is preferred. shape : tuple, default (1,) is the shape of the returned array (shape(1, ) returns a single value) for example, shape = (nlay, nrow, ncol) charlen : int, default 16 is the length character bytes. Note that arrays of bytes cannot be returned, only multi-character bytes. Shape has no affect on bytes. Raises ------ EOFError """ if vartype == str: # handle a hang-over from python2 warnings.warn( "vartype=str is deprecated; use vartype=bytes instead.", DeprecationWarning, ) vartype = bytes if vartype == bytes: # read character bytes of length charlen result = file.read(charlen) if len(result) < charlen: raise EOFError else: # find the number of values nval = np.prod(shape) result = np.fromfile(file, vartype, nval) if result.size < nval: raise EOFError if nval != 1: result = np.reshape(result, shape) return result
[docs]def join_struct_arrays(arrays): """ Simple function that can join two numpy structured arrays. """ newdtype = sum((a.dtype.descr for a in arrays), []) newrecarray = np.empty(len(arrays[0]), dtype=newdtype) for a in arrays: for name in a.dtype.names: newrecarray[name] = a[name] return newrecarray
[docs]def get_headfile_precision(filename: Union[str, PathLike]): """ Determine precision of a MODFLOW head file. Parameters ---------- filename : str or PathLike Path of binary MODFLOW file to determine precision. Returns ------- str Result will be unknown, single, or double """ # Set default result if neither single or double works result = "unknown" # Open file, and check filesize to ensure this is not an empty file f = open(filename, "rb") f.seek(0, 2) totalbytes = f.tell() f.seek(0, 0) # reset to beginning assert f.tell() == 0 if totalbytes == 0: raise ValueError(f"datafile error: file is empty: {filename}") # first try single vartype = [ ("kstp", "<i4"), ("kper", "<i4"), ("pertim", "<f4"), ("totim", "<f4"), ("text", "S16"), ] hdr = binaryread(f, vartype) charbytes = list(hdr[0][4]) if min(charbytes) >= 32 and max(charbytes) <= 126: # check if bytes are within conventional ASCII range result = "single" success = True else: success = False # next try double if not success: f.seek(0) vartype = [ ("kstp", "<i4"), ("kper", "<i4"), ("pertim", "<f8"), ("totim", "<f8"), ("text", "S16"), ] hdr = binaryread(f, vartype) charbytes = list(hdr[0][4]) if min(charbytes) >= 32 and max(charbytes) <= 126: result = "double" else: f.close() raise ValueError( f"Could not determine the precision of the headfile {filename}" ) # close and return result f.close() return result
[docs]class BinaryLayerFile(LayerFile): """ The BinaryLayerFile class is a parent class from which concrete classes inherit. This class should not be instantiated directly. Notes ----- The BinaryLayerFile class is built on a record array consisting of headers, which are record arrays of the modflow header information (kstp, kper, pertim, totim, text, nrow, ncol, ilay), and long ints pointing to the 1st byte of data for the corresponding data arrays. """ def __init__(self, filename: Union[str, PathLike], precision, verbose, **kwargs): super().__init__(filename, precision, verbose, **kwargs) def _build_index(self): """ Build the recordarray and iposarray, which maps the header information to the position in the binary file. """ header = self._get_header() self.nrow = header["nrow"] self.ncol = header["ncol"] if header["ilay"] > self.nlay: self.nlay = header["ilay"] if self.nrow < 0 or self.ncol < 0: raise ValueError("negative nrow, ncol") warn_threshold = 10000000 if self.nrow > 1 and self.nrow * self.ncol > warn_threshold: warnings.warn( f"Very large grid, ncol ({self.ncol}) * nrow ({self.nrow})" f" > {warn_threshold}" ) self.file.seek(0, 2) self.totalbytes = self.file.tell() self.file.seek(0, 0) ipos = 0 while ipos < self.totalbytes: header = self._get_header() self.recordarray.append(header) if self.text.upper() not in header["text"]: continue if ipos == 0: self.times.append(header["totim"]) self.kstpkper.append((header["kstp"], header["kper"])) else: totim = header["totim"] if totim != self.times[-1]: self.times.append(totim) self.kstpkper.append((header["kstp"], header["kper"])) ipos = self.file.tell() self.iposarray.append(ipos) databytes = self.get_databytes(header) self.file.seek(databytes, 1) ipos = self.file.tell() # self.recordarray contains a recordarray of all the headers. self.recordarray = np.array(self.recordarray, dtype=self.header_dtype) self.iposarray = np.array(self.iposarray, dtype=np.int64) self.nlay = np.max(self.recordarray["ilay"]) # provide headers as a pandas frame self.headers = pd.DataFrame(self.recordarray, index=self.iposarray) self.headers["text"] = ( self.headers["text"].str.decode("ascii", "strict").str.strip() )
[docs] def get_databytes(self, header): """ Parameters ---------- header : datafile.Header header object Returns ------- databytes : int size of the data array, in bytes, following the header """ return ( np.int64(header["ncol"]) * np.int64(header["nrow"]) * np.int64(self.realtype(1).nbytes) )
def _read_data(self, shp): return binaryread(self.file, self.realtype, shape=shp) def _get_header(self): """ Read the file header """ header = binaryread(self.file, self.header_dtype, (1,)) return header[0]
[docs] def get_ts(self, idx): """ Get a time series from the binary file. Parameters ---------- idx : int, tuple of ints, or list of such Acceptable values depend on grid type: - Structured grids (DIS): (layer, row, column) or list of such - Vertex grids (DISV): (layer, cellid) or list of such - Unstructured grids (DISU): node number or list of such All indices must be zero-based. For backwards compatibility, DISV and DISU grids also accept the old 3-tuple format with dummy values: (layer, dummy, cellid) for DISV and (dummy, dummy, node) for DISU. Returns ------- out : numpy array Array has size (ntimes, ncells + 1). The first column in the data array will contain time (totim). See Also -------- Notes ----- Index ranges (zero-based): - DIS: 0 <= layer < nlay, 0 <= row < nrow, 0 <= col < ncol - DISV: 0 <= layer < nlay, 0 <= cellid < ncpl - DISU: 0 <= node < nnodes Examples -------- >>> # DIS grid: layer 0, row 5, column 5 >>> ts = hds.get_ts(idx=(0, 5, 5)) >>> # DISV grid: layer 0, cell 12 >>> ts = hds.get_ts(idx=(0, 12)) >>> # DISU grid: node 10 >>> ts = hds.get_ts(idx=10) """ kijlist = self._build_kijlist(idx) nstation = self._get_nstation(idx, kijlist) # Initialize result array and put times in first column result = self._init_result(nstation) # Determine grid type grid_type = "structured" if self.modelgrid is None else self.modelgrid.grid_type istat = 1 for item in kijlist: # Unpack based on grid type if grid_type == "structured": # DIS: 3-tuple (k, i, j) k, i, j = item ioffset = (i * self.ncol + j) * self.realtype(1).nbytes elif grid_type == "vertex": # DISV: 2-tuple (k, cellid) k, cell = item ioffset = cell * self.realtype(1).nbytes else: # DISU: integer (node) node = item ioffset = node * self.realtype(1).nbytes k = 0 # dummy value for DISU for irec, header in enumerate(self.recordarray): ilay = header["ilay"] - 1 # change ilay from header to zero-based # For structured and vertex grids, check layer matches # For unstructured grids, read from the single "layer" in the file if grid_type != "unstructured" and ilay != k: continue ipos = self.iposarray[irec].item() # Calculate offset necessary to reach intended cell self.file.seek(ipos + int(ioffset), 0) # Find the time index and then put value into result in the # correct location. itim = np.asarray(result[:, 0] == header["totim"]).nonzero()[0] result[itim, istat] = binaryread(self.file, self.realtype) istat += 1 return result
[docs]class HeadFile(BinaryLayerFile): """ The HeadFile class provides simple ways to retrieve and manipulate 2D or 3D head arrays, or time series arrays for one or more cells, from a binary head output file. A utility method is also provided to reverse the order of head data, for use with particle tracking simulations in which particles are tracked backwards in time from terminating to release locations (e.g., to compute capture zones). Parameters ---------- filename : str or PathLike Path of the head file. text : string Name of the text string in the head file. Default is 'head'. precision : string Precision of floating point head data in the value. Accepted values are 'auto', 'single' or 'double'. Default is 'auto', which enables automatic detection of precision. verbose : bool Toggle logging output. Default is False. Examples -------- >>> import flopy.utils.binaryfile as bf >>> hdobj = bf.HeadFile('model.hds', precision='single') >>> hdobj.headers >>> rec = hdobj.get_data(kstpkper=(0, 49)) >>> ddnobj = bf.HeadFile('model.ddn', text='drawdown', precision='single') >>> ddnobj.headers >>> rec = ddnobj.get_data(totim=100.) """ def __init__( self, filename: Union[str, PathLike], text="head", precision="auto", verbose=False, **kwargs, ): self.text = text.encode() if precision == "auto": precision = get_headfile_precision(filename) if precision == "unknown": raise ValueError( f"Error. Precision could not be determined for {filename}" ) self.header_dtype = BinaryHeader.set_dtype(bintype="Head", precision=precision) super().__init__(filename, precision, verbose, **kwargs)
[docs] def reverse(self, filename: Optional[PathLike] = None): """ Reverse the time order of the currently loaded binary head file. If a head file name is not provided or the provided name is the same as the existing filename, the file will be overwritten and reloaded. Parameters ---------- filename : str or PathLike, optional Path of the reversed binary head file. """ filename = ( Path(filename).expanduser().absolute() if filename is not None else self.filename ) time = ModelTime.from_headers(self.recordarray) time._set_totim_dict() trev = time.reverse() trev._set_totim_dict() nper = time.nper seen = set() def reverse_header(header): """Reverse period, step and time fields in the record header""" nonlocal seen kper = header["kper"] - 1 kstp = header["kstp"] - 1 header = header.copy() header["kper"] = nper - kper header["kstp"] = time.nstp[kper] - kstp kper = header["kper"] - 1 kstp = header["kstp"] - 1 seen.add((kper, kstp)) header["pertim"] = trev._pertim_dict[kper, kstp] header["totim"] = trev._totim_dict[kper, kstp] return header target = filename # if rewriting the same file, write # temp file then copy it into place inplace = filename == self.filename if inplace: temp_dir_path = Path(tempfile.gettempdir()) temp_file_path = temp_dir_path / filename.name target = temp_file_path # reverse record order with open(target, "wb") as f: for i in range(len(self) - 1, -1, -1): header = self.recordarray[i].copy() header = reverse_header(header) text = header["text"] ilay = header["ilay"] kstp = header["kstp"] kper = header["kper"] pertim = header["pertim"] totim = header["totim"] data = self.get_data(idx=i)[ilay - 1] dt = np.dtype( [ ("kstp", np.int32), ("kper", np.int32), ("pertim", np.float64), ("totim", np.float64), ("text", "S16"), ("ncol", np.int32), ("nrow", np.int32), ("ilay", np.int32), ] ) nrow = data.shape[0] ncol = data.shape[1] h = np.array( (kstp, kper, pertim, totim, text, ncol, nrow, ilay), dtype=dt ) h.tofile(f) data.tofile(f) # if we rewrote the original file, reinitialize if inplace: move(target, filename) super().__init__(filename, self.precision, self.verbose)
[docs]class UcnFile(BinaryLayerFile): """ UcnFile Class. Parameters ---------- filename : string Name of the concentration file text : string Name of the text string in the ucn file. Default is 'CONCENTRATION' precision : string 'auto', 'single' or 'double'. Default is 'auto'. verbose : bool Write information to the screen. Default is False. Attributes ---------- Methods ------- See Also -------- Notes ----- The UcnFile class provides simple ways to retrieve 2d and 3d concentration arrays from a MT3D binary head file and time series arrays for one or more cells. The BinaryLayerFile class is built on a record array consisting of headers, which are record arrays of the modflow header information (kstp, kper, pertim, totim, text, nrow, ncol, ilay) and long integers, which are pointers to first bytes of data for the corresponding data array. Examples -------- >>> import flopy.utils.binaryfile as bf >>> ucnobj = bf.UcnFile('MT3D001.UCN', precision='single') >>> ucnobj.headers >>> rec = ucnobj.get_data(kstpkper=(0, 0)) """ def __init__( self, filename, text="concentration", precision="auto", verbose=False, **kwargs, ): self.text = text.encode() if precision == "auto": precision = get_headfile_precision(filename) if precision == "unknown": raise ValueError(f"Error. Precision could not be determined for {filename}") self.header_dtype = BinaryHeader.set_dtype(bintype="Ucn", precision=precision) super().__init__(filename, precision, verbose, **kwargs) return
[docs]class HeadUFile(BinaryLayerFile): """ The HeadUFile class provides simple ways to retrieve a list of head arrays from a MODFLOW-USG binary head file and time series arrays for one or more cells. Parameters ---------- filename : str or PathLike Path of the head file text : string Name of the text string in the head file. Default is 'headu'. precision : string Precision of the floating point head data in the file. Accepted values are 'auto', 'single' or 'double'. Default is 'auto', which enables precision to be automatically detected. verbose : bool Toggle logging output. Default is False. Notes ----- The BinaryLayerFile class is built on a record array consisting of headers, which are record arrays of the modflow header information (kstp, kper, pertim, totim, text, nrow, ncol, ilay), and long ints pointing to the 1st byte of data for the corresponding data arrays. This class overrides methods in the parent class so that the proper sized arrays are created: for unstructured grids, nrow and ncol are the starting and ending node numbers for layer, ilay. When the get_data method is called for this class, a list of one-dimensional arrays will be returned, where each array is the head array for a layer. If the heads for a layer were not saved, then None will be returned for that layer. Examples -------- >>> import flopy.utils.binaryfile as bf >>> hdobj = bf.HeadUFile('model.hds') >>> hdobj.headers >>> usgheads = hdobj.get_data(kstpkper=(0, 49)) """ def __init__( self, filename: Union[str, PathLike], text="headu", precision="auto", verbose=False, **kwargs, ): """ Class constructor """ self.text = text.encode() if precision == "auto": precision = get_headfile_precision(filename) if precision == "unknown": raise ValueError( f"Error. Precision could not be determined for {filename}" ) self.header_dtype = BinaryHeader.set_dtype(bintype="Head", precision=precision) super().__init__(filename, precision, verbose, **kwargs) def _get_data_array(self, totim=0.0): """ Get a list of 1D arrays for the specified kstp and kper value or totim value. """ if totim >= 0.0: keyindices = np.asarray(self.recordarray["totim"] == totim).nonzero()[0] if len(keyindices) == 0: raise ValueError(f"totim value ({totim}) not found in file") else: raise ValueError("Data not found") # fill a list of 1d arrays with heads from binary file data = self.nlay * [None] for idx in keyindices: ipos = self.iposarray[idx] ilay = self.recordarray["ilay"][idx] nstrt = self.recordarray["ncol"][idx] nend = self.recordarray["nrow"][idx] npl = nend - nstrt + 1 if self.verbose: print(f"Byte position in file: {ipos} for layer {ilay}") self.file.seek(ipos, 0) data[ilay - 1] = binaryread(self.file, self.realtype, shape=(npl,)) return data
[docs] def get_databytes(self, header): """ Parameters ---------- header : datafile.Header header object Returns ------- databytes : int size of the data array, in bytes, following the header """ # unstructured head files contain node starting and ending indices # for each layer nstrt = np.int64(header["ncol"]) nend = np.int64(header["nrow"]) npl = nend - nstrt + 1 return npl * np.int64(self.realtype(1).nbytes)
[docs] def get_ts(self, idx): """ Get a time series from the binary HeadUFile Parameters ---------- idx : int or list of ints idx can be nodenumber or it can be a list in the form [nodenumber, nodenumber, ...]. The nodenumber, values must be zero based. Returns ------- out : numpy array Array has size (ntimes, ncells + 1). The first column in the data array will contain time (totim). """ times = self.get_times() data = self.get_data(totim=times[0]) layers = len(data) ncpl = [len(data[l]) for l in range(layers)] result = [] if isinstance(idx, int): layer, nn = get_lni(ncpl, [idx])[0] for i, time in enumerate(times): data = self.get_data(totim=time) value = data[layer][nn] result.append([time, value]) elif isinstance(idx, list) and all(isinstance(x, int) for x in idx): for i, time in enumerate(times): data = self.get_data(totim=time) row = [time] lni = get_lni(ncpl, idx) for layer, nn in lni: value = data[layer][nn] row += [value] result.append(row) else: raise ValueError("idx must be an integer or a list of integers") return np.array(result)
[docs] def get_alldata(self, mflay=None, nodata=-9999): """ Get all data from the USG head file. Parameters ---------- mflay : integer MODFLOW zero-based layer number to return. For USG files, this parameter is required (cannot be None). (Default is None.) nodata : float The nodata value in the data array. All array values that have the nodata value will be assigned np.nan. Returns ------- data : numpy array Array has size (ntimes, ncells_in_layer) when mflay is specified. Raises ------ NotImplementedError Raised when mflay=None. USG head files contain ragged arrays with variable-sized data per layer, which cannot be converted to a uniform numpy array when retrieving all layers. Notes ----- For MODFLOW-USG files, the mflay parameter must be specified to retrieve data for a single layer across all timesteps. To get all layers for a specific timestep, use get_data() instead. """ if mflay is None: raise NotImplementedError( "get_alldata() with mflay=None is not supported for" "MODFLOW-USG head files. These contain variably-size " "data per layer which cannot be stacked into a numpy " "array. Specify mflay to get data for a single layer " "or use get_data() for specific timesteps." ) return super().get_alldata(mflay=mflay, nodata=nodata)
[docs]class BudgetIndexError(Exception): pass
[docs]class CellBudgetFile: """ The CellBudgetFile class provides convenient ways to retrieve and manipulate budget data from a binary cell budget file. A utility method is also provided to reverse the budget records for particle tracking simulations in which particles are tracked backwards from terminating to release locations (e.g., to compute capture zones). Parameters ---------- filename : str or PathLike Path of the cell budget file. precision : string Precision of floating point budget data in the file. Accepted values are 'single' or 'double'. Default is 'single'. verbose : bool Toggle logging output. Default is False. Examples -------- >>> import flopy.utils.binaryfile as bf >>> cbb = bf.CellBudgetFile('mymodel.cbb') >>> cbb.headers >>> rec = cbb.get_data(kstpkper=(0,0), text='RIVER LEAKAGE') """ def __init__( self, filename: Union[str, PathLike], precision="auto", verbose=False, **kwargs, ): self.filename = Path(filename).expanduser().absolute() self.precision = precision self.verbose = verbose self.file = open(self.filename, "rb") # Get filesize to ensure this is not an empty file self.file.seek(0, 2) totalbytes = self.file.tell() self.file.seek(0, 0) # reset to beginning assert self.file.tell() == 0 if totalbytes == 0: raise ValueError(f"datafile error: file is empty: {filename}") self.nrow = 0 self.ncol = 0 self.nlay = 0 self.nper = 0 self.times = [] self.kstpkper = [] self.recordarray = [] self.iposheader = [] self.iposarray = [] self.textlist = [] self.imethlist = [] self.paknamlist_from = [] self.paknamlist_to = [] self.compact = True # compact budget file flag self.dis = None self.modelgrid = None if "model" in kwargs.keys(): self.model = kwargs.pop("model") self.modelgrid = self.model.modelgrid self.dis = self.model.dis if "dis" in kwargs.keys(): self.dis = kwargs.pop("dis") self.modelgrid = self.dis.parent.modelgrid if "tdis" in kwargs.keys(): self.tdis = kwargs.pop("tdis") if "modelgrid" in kwargs.keys(): self.modelgrid = kwargs.pop("modelgrid") if len(kwargs.keys()) > 0: args = ",".join(kwargs.keys()) raise ValueError(f"LayerFile error: unrecognized kwargs: {args}") if precision == "auto": success = self._set_precision("single") if not success: success = self._set_precision("double") if not success: s = "Budget precision could not be auto determined" raise BudgetIndexError(s) elif precision == "single": success = self._set_precision(precision) elif precision == "double": success = self._set_precision(precision) else: raise ValueError(f"Unknown precision specified: {precision}") # set shape for full3D option if self.modelgrid is None: self.shape = (self.nlay, self.nrow, self.ncol) self.nnodes = self.nlay * self.nrow * self.ncol else: self.shape = self.modelgrid.shape self.nnodes = self.modelgrid.nnodes if not success: raise ValueError( f"Budget file could not be read using {precision} precision" ) def __enter__(self): return self def __exit__(self, *exc): self.close() def __len__(self): """ Return the number of records (headers) in the file. """ return len(self.recordarray) @property def nrecords(self): """ Return the number of records (headers) in the file. .. deprecated:: 3.8.0 Use :meth:`len` instead. """ warnings.warn( "obj.nrecords is deprecated; use len(obj) instead.", DeprecationWarning, ) return len(self) def __reset(self): """ Reset indexing lists when determining precision """ self.file.seek(0, 0) self.times = [] self.kstpkper = [] self.recordarray = [] self.iposheader = [] self.iposarray = [] self.textlist = [] self.imethlist = [] self.paknamlist_from = [] self.paknamlist_to = [] def _set_precision(self, precision="single"): """ Method to set the budget precision from a CBC file. Enables Auto precision code to work Parameters ---------- precision : str budget file precision (accepts 'single' or 'double') """ success = True h1dt = [ ("kstp", "i4"), ("kper", "i4"), ("text", "S16"), ("ncol", "i4"), ("nrow", "i4"), ("nlay", "i4"), ] if precision == "single": self.realtype = np.float32 ffmt = "f4" else: self.realtype = np.float64 ffmt = "f8" h2dt0 = [ ("imeth", "i4"), ("delt", ffmt), ("pertim", ffmt), ("totim", ffmt), ] h2dt = [ ("imeth", "i4"), ("delt", ffmt), ("pertim", ffmt), ("totim", ffmt), ("modelnam", "S16"), ("paknam", "S16"), ("modelnam2", "S16"), ("paknam2", "S16"), ] self.header1_dtype = np.dtype(h1dt) self.header2_dtype0 = np.dtype(h2dt0) self.header2_dtype = np.dtype(h2dt) hdt = h1dt + h2dt self.header_dtype = np.dtype(hdt) try: self._build_index() except (BudgetIndexError, EOFError) as e: success = False self.__reset() return success def _totim_from_kstpkper(self, kstpkper): if self.dis is None: return -1.0 kstp, kper = kstpkper perlen = self.dis.perlen.array nstp = self.dis.nstp.array[kper] tsmult = self.dis.tsmult.array[kper] kper_len = np.sum(perlen[:kper]) this_perlen = perlen[kper] if tsmult == 1: dt1 = this_perlen / float(nstp) else: dt1 = this_perlen * (tsmult - 1.0) / ((tsmult**nstp) - 1.0) kstp_len = [dt1] for i in range(kstp + 1): kstp_len.append(kstp_len[-1] * tsmult) kstp_len = sum(kstp_len[: kstp + 1]) return kper_len + kstp_len def _build_index(self): """ Build the ordered dictionary, which maps the header information to the position in the binary file. """ # read first record header = self._get_header() nrow = header["nrow"] ncol = header["ncol"] text = header["text"].decode("ascii").strip() if nrow < 0 or ncol < 0: raise ValueError("negative nrow, ncol") if text != "FLOW-JA-FACE": self.nrow = nrow self.ncol = ncol self.nlay = np.abs(header["nlay"]) self.file.seek(0, 2) self.totalbytes = self.file.tell() self.file.seek(0, 0) self.recorddict = {} # read the remaining records ipos = 0 while ipos < self.totalbytes: self.iposheader.append(ipos) header = self._get_header() totim = header["totim"] # if old-style (non-compact) file, # compute totim from kstp and kper if not self.compact: totim = self._totim_from_kstpkper( (header["kstp"] - 1, header["kper"] - 1) ) header["totim"] = totim if totim >= 0 and totim not in self.times: self.times.append(totim) kstpkper = (header["kstp"], header["kper"]) if kstpkper not in self.kstpkper: self.kstpkper.append(kstpkper) if header["text"] not in self.textlist: # check the precision of the file using text records tlist = [header["text"], header["modelnam"]] for text in tlist: if len(text) == 0: continue charbytes = list(text) if min(charbytes) < 32 or max(charbytes) > 126: # not in conventional ASCII range raise BudgetIndexError("Improper precision") self.textlist.append(header["text"]) self.imethlist.append(header["imeth"]) if header["paknam"] not in self.paknamlist_from: self.paknamlist_from.append(header["paknam"]) if header["paknam2"] not in self.paknamlist_to: self.paknamlist_to.append(header["paknam2"]) ipos = self.file.tell() if self.verbose: for itxt in [ "kstp", "kper", "text", "ncol", "nrow", "nlay", "imeth", "delt", "pertim", "totim", "modelnam", "paknam", "modelnam2", "paknam2", ]: s = header[itxt] print(f"{itxt}: {s}") print("file position: ", ipos) if header["imeth"].item() not in {5, 6, 7}: print("") # set the nrow, ncol, and nlay if they have not been set if self.nrow == 0: text = header["text"].decode("ascii").strip() if text != "FLOW-JA-FACE": self.nrow = header["nrow"] self.ncol = header["ncol"] self.nlay = np.abs(header["nlay"]) # store record and byte position mapping self.recorddict[tuple(header)] = ( ipos # store the position right after header2 ) self.recordarray.append(header) self.iposarray.append(ipos) # store the position right after header2 # skip over the data to the next record and set ipos self._skip_record(header) ipos = self.file.tell() # convert to numpy arrays self.recordarray = np.array(self.recordarray, dtype=self.header_dtype) self.iposheader = np.array(self.iposheader, dtype=np.int64) self.iposarray = np.array(self.iposarray, dtype=np.int64) self.nper = self.recordarray["kper"].max() # provide headers as a pandas frame self.headers = pd.DataFrame(self.recordarray, index=self.iposarray) # remove irrelevant columns cols = self.headers.columns.to_list() unique_imeth = self.headers["imeth"].unique() if unique_imeth.max() == 0: drop_cols = cols[cols.index("imeth") :] elif 6 not in unique_imeth: drop_cols = cols[cols.index("modelnam") :] else: drop_cols = [] if drop_cols: self.headers.drop(columns=drop_cols, inplace=True) for name in self.headers.columns: dtype = self.header_dtype[name] if np.issubdtype(dtype, bytes): # convert to str self.headers[name] = ( self.headers[name].str.decode("ascii", "strict").str.strip() ) def _skip_record(self, header): """ Skip over this record, not counting header and header2. """ nlay = abs(header["nlay"]) nrow = header["nrow"] ncol = header["ncol"] imeth = header["imeth"] realtype_nbytes = self.realtype(1).nbytes if imeth == 0: nbytes = nrow * ncol * nlay * realtype_nbytes elif imeth == 1: nbytes = nrow * ncol * nlay * realtype_nbytes elif imeth == 2: nlist = binaryread(self.file, np.int32)[0] nbytes = nlist * (4 + realtype_nbytes) elif imeth == 3: nbytes = nrow * ncol * realtype_nbytes + (nrow * ncol * 4) elif imeth == 4: nbytes = nrow * ncol * realtype_nbytes elif imeth == 5: nauxp1 = binaryread(self.file, np.int32)[0] naux = nauxp1 - 1 naux_nbytes = naux * 16 if naux_nbytes: check = self.file.seek(naux_nbytes, 1) if check < naux_nbytes: raise EOFError nlist = binaryread(self.file, np.int32)[0] if self.verbose: print("naux: ", naux) print("nlist: ", nlist) print("") nbytes = nlist * (4 + realtype_nbytes + naux * realtype_nbytes) elif imeth == 6: # read rest of list data nauxp1 = binaryread(self.file, np.int32)[0] naux = nauxp1 - 1 naux_nbytes = naux * 16 if naux_nbytes: check = self.file.seek(naux_nbytes, 1) if check < naux_nbytes: raise EOFError nlist = binaryread(self.file, np.int32)[0] if self.verbose: print("naux: ", naux) print("nlist: ", nlist) print("") nbytes = nlist * (4 * 2 + realtype_nbytes + naux * realtype_nbytes) else: raise ValueError(f"invalid method code {imeth}") if nbytes != 0: self.file.seek(nbytes, 1) def _get_header(self): """ Read the file header """ header1 = binaryread(self.file, self.header1_dtype, (1,)) nlay = header1["nlay"] self.compact = bool(nlay < 0) if self.compact: # fill header2 by first reading imeth, delt, pertim and totim # and then adding modelnames and paknames if imeth = 6 temp = binaryread(self.file, self.header2_dtype0, (1,)) header2 = np.array( [(0, 0.0, 0.0, 0.0, "", "", "", "")], dtype=self.header2_dtype ) for name in temp.dtype.names: header2[name] = temp[name] if header2["imeth"].item() == 6: header2["modelnam"] = binaryread(self.file, bytes, charlen=16) header2["paknam"] = binaryread(self.file, bytes, charlen=16) header2["modelnam2"] = binaryread(self.file, bytes, charlen=16) header2["paknam2"] = binaryread(self.file, bytes, charlen=16) else: header2 = np.array( [(0, 0.0, 0.0, 0.0, "", "", "", "")], dtype=self.header2_dtype ) fullheader = join_struct_arrays([header1, header2]) return fullheader[0] def _find_text(self, text): """ Determine if selected record name is in budget file """ # check and make sure that text is in file text16 = None if text is not None: if isinstance(text, bytes): ttext = text.decode() else: ttext = text for t in self.textlist: if ttext.upper() in t.decode(): text16 = t break if text16 is None: raise ValueError("The specified text string is not in the budget file") return text16 def _find_paknam(self, paknam, to=False): """ Determine if selected record name is in budget file """ # check and make sure that text is in file paknam16 = None if paknam is not None: if isinstance(paknam, bytes): tpaknam = paknam.decode() else: tpaknam = paknam for t in self._unique_package_names(to): if tpaknam.upper() in t.decode(): paknam16 = t break if paknam16 is None: raise ValueError( "The specified package name string is not in the budget file" ) return paknam16
[docs] def list_records(self): """ Print a list of all of the records in the file .. deprecated:: 3.8.0 Use :attr:`headers` instead. """ warnings.warn( "list_records() is deprecated; use headers instead.", DeprecationWarning, ) for rec in self.recordarray: if isinstance(rec, bytes): rec = rec.decode() print(rec)
[docs] def list_unique_records(self): """ Print a list of unique record names .. deprecated:: 3.8.0 Use `headers[["text", "imeth"]].drop_duplicates()` instead. """ warnings.warn( "list_unique_records() is deprecated; use " 'headers[["text", "imeth"]].drop_duplicates() instead.', DeprecationWarning, ) print("RECORD IMETH") print(22 * "-") for rec, imeth in zip(self.textlist, self.imethlist): if isinstance(rec, bytes): rec = rec.decode() print(f"{rec.strip():16} {imeth:5d}")
[docs] def list_unique_packages(self, to=False): """ Print a list of unique package names .. deprecated:: 3.8.0 Use `headers.paknam.drop_duplicates()` or `headers.paknam2.drop_duplicates()` instead. """ warnings.warn( "list_unique_packages() is deprecated; use " "headers.paknam.drop_duplicates() or " "headers.paknam2.drop_duplicates() instead", DeprecationWarning, ) for rec in self._unique_package_names(to): if isinstance(rec, bytes): rec = rec.decode() print(rec)
[docs] def get_unique_record_names(self, decode=False): """ Get a list of unique record names in the file Parameters ---------- decode : bool Optional boolean used to decode byte strings (default is False). Returns ------- names : list of strings List of unique text names in the binary file. """ if decode: names = [] for text in self.textlist: if isinstance(text, bytes): text = text.decode() names.append(text) else: names = self.textlist return names
[docs] def get_unique_package_names(self, decode=False, to=False): """ Get a list of unique package names in the file Parameters ---------- decode : bool Optional boolean used to decode byte strings (default is False). Returns ------- names : list of strings List of unique package names in the binary file. """ if decode: names = [] for text in self._unique_package_names(to): if isinstance(text, bytes): text = text.decode() names.append(text) else: names = self._unique_package_names(to) return names
def _unique_package_names(self, to=False): """ Get a list of unique package names in the file Returns ------- out : list of strings List of unique package names in the binary file. """ return self.paknamlist_to if to else self.paknamlist_from
[docs] def get_kstpkper(self): """ Get a list of unique tuples (stress period, time step) in the file. Indices are 0-based, use the `kstpkper` attribute for 1-based. Returns ------- list of (kstp, kper) tuples List of unique combinations of stress period & time step indices (0-based) in the binary file """ return [(kstp - 1, kper - 1) for kstp, kper in self.kstpkper]
[docs] def get_indices(self, text=None): """ Get a list of indices for a selected record name Parameters ---------- text : str The text identifier for the record. Examples include 'RIVER LEAKAGE', 'STORAGE', 'FLOW RIGHT FACE', etc. Returns ------- out : tuple indices of selected record name in budget file. """ # check and make sure that text is in file if text is not None: text16 = self._find_text(text) select_indices = np.asarray(self.recordarray["text"] == text16).nonzero() if isinstance(select_indices, tuple): select_indices = select_indices[0] else: select_indices = None return select_indices
[docs] def get_position(self, idx, header=False): """ Get the starting position of the data or header for a specified record number in the binary budget file. Parameters ---------- idx : int The zero-based record number. The first record is record 0. header : bool If True, the position of the start of the header data is returned. If False, the position of the start of the data is returned (default is False). Returns ------- ipos : int64 The position of the start of the data in the cell budget file or the start of the header. """ if header: ipos = self.iposheader[idx] else: ipos = self.iposarray[idx] return ipos
[docs] def get_data( self, idx=None, kstpkper=None, totim=None, text=None, paknam=None, paknam2=None, full3D=False, ) -> Union[list, np.ndarray]: """ Get data from the binary budget file. Parameters ---------- idx : int or list The zero-based record number. The first record is record 0. kstpkper : tuple of ints A tuple containing the time step and stress period (kstp, kper). The kstp and kper values are zero based. totim : float The simulation time. text : str The text identifier for the record. Examples include 'RIVER LEAKAGE', 'STORAGE', 'FLOW RIGHT FACE', etc. paknam : str The `from` package name for the record. paknam2 : str The `to` package name for the record. This argument can be useful for MODFLOW 6 budget files if multiple packages of the same type are specified. The paknam2 argument can be specified as the package name (not the package type) in order to retrieve budget data for a specific named package. full3D : boolean If true, then return the record as a three dimensional numpy array, even for those list-style records written as part of a 'COMPACT BUDGET' MODFLOW budget file. (Default is False.) Returns ------- recordlist : list of records A list of budget objects. The structure of the returned object depends on the structure of the data in the cbb file. If full3D is True, then this method will return a numpy masked array of size (nlay, nrow, ncol) for those list-style 'COMPACT BUDGET' records written by MODFLOW. See Also -------- Notes ----- Examples -------- """ # trap for totim error if totim is not None: if len(self.times) == 0: errmsg = """This is an older style budget file that does not have times in it. Use the MODFLOW compact budget format if you want to work with times. Or you may access this file using the kstp and kper arguments or the idx argument.""" raise ValueError(errmsg) # check and make sure that text is in file text16 = None if text is not None: text16 = self._find_text(text) paknam16 = None if paknam is not None: paknam16 = self._find_paknam(paknam) paknam16_2 = None if paknam2 is not None: paknam16_2 = self._find_paknam(paknam2, to=True) # build the selection mask select_indices = np.array([True] * len(self.recordarray)) selected = False if idx is not None: select_indices[idx] = False select_indices = ~select_indices selected = True if kstpkper is not None: kstp1 = kstpkper[0] + 1 kper1 = kstpkper[1] + 1 select_indices = select_indices & (self.recordarray["kstp"] == kstp1) select_indices = select_indices & (self.recordarray["kper"] == kper1) selected = True if text16 is not None: select_indices = select_indices & (self.recordarray["text"] == text16) selected = True if paknam16 is not None: select_indices = select_indices & (self.recordarray["paknam"] == paknam16) selected = True if paknam16_2 is not None: select_indices = select_indices & ( self.recordarray["paknam2"] == paknam16_2 ) selected = True if totim is not None: select_indices = select_indices & np.isclose( self.recordarray["totim"], totim ) selected = True if not selected: raise TypeError( "get_data() missing 1 required argument: 'kstpkper', 'totim', " "'idx', or 'text'" ) return [ self.get_record(idx, full3D=full3D) for idx, t in enumerate(select_indices) if t ]
[docs] def to_geodataframe( self, gdf=None, modelgrid=None, idx=None, kstpkper=None, totim=None, text=None, ): if (idx, kstpkper, totim) == (None, None, None): raise AssertionError( "to_geodataframe() missing 1 required argument: " "please provide 'idx', 'kstpkper', or 'totim'" ) if gdf is None: if modelgrid is None: if self.modelgrid is None: raise AssertionError( "A geodataframe or modelgrid instance must be supplied" ) modelgrid = self.modelgrid gdf = modelgrid.to_geodataframe() col_names = [] if text is None: textlist = [i.decode() for i in self.textlist] for text in textlist: data = self.get_data( idx=idx, kstpkper=kstpkper, totim=totim, text=text, full3D=True ) text = text.strip().lower().replace(" ", "_") for ix, arr in enumerate(data[0]): name = f"{text}_{ix}" gdf[name] = arr.ravel() col_names.append(name) else: data = self.get_data( idx=idx, kstpkper=kstpkper, totim=totim, text=text, full3D=True ) text = text.strip().lower().replace(" ", "_") for ix, arr in enumerate(data[0]): name = f"{text}_{ix}" gdf[name] = arr.ravel() col_names.append(name) return gdf
[docs] def get_ts(self, idx, text=None, times=None, variable="q"): """ Get a time series from the binary budget file. Parameters ---------- idx : int, tuple of ints, or list of such Acceptable values depend on grid type: - Structured grids (DIS): (layer, row, column) or list of such - Vertex grids (DISV): (layer, cellid) or list of such - Unstructured grids (DISU): node number or list of such All indices must be zero-based. For backwards compatibility, DISV and DISU grids also accept the old 3-tuple format with dummy values: (layer, dummy, cellid) for DISV and (dummy, dummy, node) for DISU. text : str The text identifier for the record. Examples include 'RIVER LEAKAGE', 'STORAGE', 'FLOW RIGHT FACE', etc. times : iterable of floats List of times to from which to get time series. variable : str, optional Variable name to extract from the budget record. Default is 'q'. For records with auxiliary variables (e.g., 'DATA-SPDIS', 'DATA-SAT'), this can be used to access auxiliary data. Examples include 'qx', 'qy', 'qz' for specific discharge components. Returns ------- out : numpy array Array has size (ntimes, ncells + 1). The first column in the data array will contain time (totim). Notes ----- Index ranges (zero-based): - DIS: 0 <= layer < nlay, 0 <= row < nrow, 0 <= col < ncol - DISV: 0 <= layer < nlay, 0 <= cellid < ncpl - DISU: 0 <= node < nnodes Examples -------- >>> # DIS grid: layer 0, row 5, column 5 >>> ts = cbb.get_ts(idx=(0, 5, 5), text='DATA-SPDIS', variable='qx') >>> # DISV grid: layer 0, cell 12 >>> ts = cbb.get_ts(idx=(0, 12), text='DATA-SPDIS', variable='qx') >>> # DISU grid: node 10 >>> ts = cbb.get_ts(idx=10, text='FLOW-JA-FACE') """ if text is None: raise ValueError( "text keyword must be provided to CellBudgetFile get_ts() method." ) cellids = self._cellids(idx) ncells = len(cellids) result = np.empty((len(self.kstpkper), ncells + 1), dtype=self.realtype) result[:, :] = np.nan if len(self.times) == result.shape[0]: result[:, 0] = np.array(self.times) timesint = self.get_times() kstpkper = self.get_kstpkper() nsteps = len(kstpkper) if len(timesint) < 1: if times is None: timesint = [x + 1 for x in range(nsteps)] else: if isinstance(times, np.ndarray): times = times.tolist() if len(times) != nsteps: raise ValueError( f"number of times provided ({len(times)}) must equal " f"number of time steps in cell budget file ({nsteps})" ) timesint = times for i, t in enumerate(timesint): result[i, 0] = t use_full3d = variable == "q" and ( self.modelgrid is None or self.modelgrid.grid_type == "structured" ) for itim, kstpkper_ in enumerate(kstpkper): if use_full3d: v = self.get_data(kstpkper=kstpkper_, text=text, full3D=True) if len(v) == 0: continue v = v[0] istat = 1 for k, i, j in cellids: result[itim, istat] = v[k, i, j].copy() istat += 1 else: v = self.get_data(kstpkper=kstpkper_, text=text) if len(v) == 0: continue if self.modelgrid is None: raise ValueError( "A modelgrid instance must be provided during " "instantiation to get IMETH=6 timeseries data" ) nodes = self._cellid_to_node(cellids) for vv in v: # Check if this is a recarray (IMETH=6) or plain array (IMETH=1) if vv.dtype.names is None: # IMETH=1: Plain array - extract values at specified cells # MODFLOW 6 stores IMETH=1 data as 3D arrays with padding: # - DIS: (nlay, nrow, ncol) - natural 3D structure # - DISV: (nlay, 1, ncpl) - middle dim is padding # - DISU: (1, 1, nnodes) - first two dims are padding if self.modelgrid.grid_type == "vertex": # DISV: shape is (nlay, 1, ncpl) # Extract as vv[k, 0, cellid] to handle padding dimension istat = 1 for k, cell in cellids: result[itim, istat] = vv[k, 0, cell].copy() istat += 1 else: # DISU: shape is (1, 1, nnodes) # Extract as vv[0, 0, node] to handle padding dimensions istat = 1 for node in cellids: result[itim, istat] = vv[0, 0, node].copy() istat += 1 else: # IMETH=6: Recarray with named fields available = vv.dtype.names if variable not in available: raise ValueError( f"Variable '{variable}' not found in budget record. " f"Available variables: {list(available)}" ) dix = np.asarray(np.isin(vv["node"], nodes)).nonzero()[0] if len(dix) > 0: result[itim, 1:] = vv[variable][dix] return result
def _cellids(self, idx): if isinstance(idx, int): idx_list = [idx] elif isinstance(idx, tuple): idx_list = [idx] elif isinstance(idx, list): idx_list = idx else: raise TypeError( f"Invalid index type, expected int, tuple " f"or list of such, got {type(idx)}" ) grid_type = "structured" if self.modelgrid is None else self.modelgrid.grid_type cellid = [] for item in idx_list: if grid_type == "structured": if not isinstance(item, (list, tuple)) or len(item) != 3: raise ValueError( f"Expected DIS cell index (layer, row, col), got: {item}" ) k, i, j = item if k < 0 or k >= self.nlay: raise ValueError(f"Layer index {k} out of range [0, {self.nlay})") if i < 0 or i >= self.nrow: raise ValueError(f"Row index {i} out of range [0, {self.nrow})") if j < 0 or j >= self.ncol: raise ValueError(f"Column index {j} out of range [0, {self.ncol})") cellid.append((k, i, j)) elif grid_type == "vertex": if isinstance(item, (list, tuple)): if len(item) == 2: # proper format: (layer, cellid) k, cell = item elif len(item) == 3: # old format: (layer, dummy, cellid) k, cell = item[0], item[2] else: raise ValueError( f"Expected DISV cell index (layer, cellid) " f"or (layer, dummy, cellid), got: {item}" ) else: raise ValueError( f"Expected DISV cell index (layer, cellid) or " f"(layer, dummy, cellid), got: {item}" ) if k < 0 or k >= self.nlay: raise ValueError(f"Layer index {k} out of range [0, {self.nlay})") if cell < 0 or cell >= self.modelgrid.ncpl: raise ValueError( f"Cell index {cell} out of range [0, {self.modelgrid.ncpl})" ) cellid.append((k, cell)) else: if isinstance(item, (int, np.integer)): # proper format: just node number node = int(item) elif isinstance(item, (list, tuple)): if len(item) == 3: # old format: (dummy, dummy, node) node = int(item[2]) elif len(item) == 1: # Also support single-element tuple node = int(item[0]) else: raise ValueError( f"Expected DISU node number or (dummy, dummy, node), " f"got: {item}" ) else: raise ValueError( f"Expected DISU node number or (dummy, dummy, node), " f"got: {item}" ) if node < 0 or node >= self.modelgrid.nnodes: raise ValueError( f"Node index {node} out of range [0, {self.modelgrid.nnodes})" ) cellid.append(node) return cellid def _cellid_to_node(self, cellids) -> list[int]: """Convert 0-based cellids to 1-based MODFLOW node numbers.""" # Get 0-based nodes from grid, then convert to 1-based (vectorized) # UnstructuredGrid.get_node() accepts both plain ints and tuples nodes_0based = self.modelgrid.get_node(cellids) return (np.array(nodes_0based) + 1).tolist()
[docs] def get_record(self, idx, full3D=False): """ Get a single data record from the budget file. Parameters ---------- idx : int The zero-based record number. The first record is record 0. full3D : boolean If true, then return the record as a three dimensional numpy array, even for those list-style records written as part of a 'COMPACT BUDGET' MODFLOW budget file. (Default is False.) Returns ------- record : a single data record The structure of the returned object depends on the structure of the data in the cbb file. Compact list data are returned as If full3D is True, then this method will return a numpy masked array of size (nlay, nrow, ncol) for those list-style 'COMPACT BUDGET' records written by MODFLOW. See Also -------- Notes ----- Examples -------- """ # idx must be an ndarray, so if it comes in as an integer then convert if np.isscalar(idx): idx = np.array([idx]) header = self.recordarray[idx] ipos = self.iposarray[idx].item() self.file.seek(ipos, 0) imeth = header["imeth"][0] t = header["text"][0].decode("ascii") s = f"Returning {t.strip()} as " nlay = abs(header["nlay"][0]) nrow = header["nrow"][0] ncol = header["ncol"][0] # default method if imeth == 0: if self.verbose: s += f"an array of shape {(nlay, nrow, ncol)}" print(s) return binaryread(self.file, self.realtype(1), shape=(nlay, nrow, ncol)) # imeth 1 elif imeth == 1: if self.verbose: s += f"an array of shape {(nlay, nrow, ncol)}" print(s) return binaryread(self.file, self.realtype(1), shape=(nlay, nrow, ncol)) # imeth 2 elif imeth == 2: nlist = binaryread(self.file, np.int32)[0] dtype = np.dtype([("node", np.int32), ("q", self.realtype)]) if self.verbose: if full3D: s += f"a numpy masked array of size ({nlay}, {nrow}, {ncol})" else: s += f"a numpy recarray of size ({nlist}, 2)" print(s) data = binaryread(self.file, dtype, shape=(nlist,)) if full3D: return self.__create3D(data) else: return data.view(np.recarray) # imeth 3 elif imeth == 3: ilayer = binaryread(self.file, np.int32, shape=(nrow, ncol)) data = binaryread(self.file, self.realtype(1), shape=(nrow, ncol)) if self.verbose: if full3D: s += f"a numpy masked array of size ({nlay}, {nrow}, {ncol})" else: s += ( "a list of two 2D numpy arrays. The first is an " f"integer layer array of shape ({nrow}, {ncol}). The " f"second is real data array of shape ({nrow}, {ncol})" ) print(s) if full3D: out = np.ma.zeros(self.nnodes, dtype=np.float32) out.mask = True vertical_layer = ilayer.flatten() - 1 # create the 2D cell index and then move it to # the correct vertical location idx = np.arange(0, vertical_layer.shape[0]) idx += vertical_layer * nrow * ncol out[idx] = data.flatten() return out.reshape(self.shape) else: return [ilayer, data] # imeth 4 elif imeth == 4: if self.verbose: s += f"a 2d numpy array of size ({nrow}, {ncol})" print(s) return binaryread(self.file, self.realtype(1), shape=(nrow, ncol)) # imeth 5 elif imeth == 5: nauxp1 = binaryread(self.file, np.int32)[0] naux = nauxp1 - 1 l = [("node", np.int32), ("q", self.realtype)] for i in range(naux): auxname = binaryread(self.file, bytes, charlen=16) l.append((auxname.decode("ascii").strip(), self.realtype)) dtype = np.dtype(l) nlist = binaryread(self.file, np.int32)[0] data = binaryread(self.file, dtype, shape=(nlist,)) if full3D: if self.verbose: s += f"a list array of shape ({nlay}, {nrow}, {ncol})" print(s) return self.__create3D(data) else: if self.verbose: s += f"a numpy recarray of size ({nlist}, {2 + naux})" print(s) return data.view(np.recarray) # imeth 6 elif imeth == 6: # read rest of list data nauxp1 = binaryread(self.file, np.int32)[0] naux = nauxp1 - 1 l = [("node", np.int32), ("node2", np.int32), ("q", self.realtype)] for i in range(naux): auxname = binaryread(self.file, bytes, charlen=16) l.append((auxname.decode("ascii").strip(), self.realtype)) dtype = np.dtype(l) nlist = binaryread(self.file, np.int32)[0] data = binaryread(self.file, dtype, shape=(nlist,)) if self.verbose: if full3D: s += f"a list array of shape ({nlay}, {nrow}, {ncol})" else: s += f"a numpy recarray of size ({nlist}, 2)" print(s) if full3D: data = self.__create3D(data) if self.modelgrid is not None: return np.reshape(data, self.shape) else: return data else: return data.view(np.recarray) else: raise ValueError(f"invalid imeth value - {imeth}") # should not reach this point return
def __create3D(self, data): """ Convert a dictionary of {node: q, ...} into a numpy masked array. Used to create full grid arrays when the full3D keyword is set to True in get_data. Parameters ---------- data : dictionary Dictionary with node keywords and flows (q) items. Returns ------- out : numpy masked array List contains unique simulation times (totim) in binary file. """ out = np.ma.zeros(self.nnodes, dtype=data["q"].dtype) out.mask = True for [node, q] in zip(data["node"], data["q"]): idx = node - 1 out.data[idx] += q out.mask[idx] = False return np.ma.reshape(out, self.shape)
[docs] def get_times(self): """ Get a list of unique times in the file Returns ------- out : list of floats List contains unique simulation times (totim) in binary file. """ return self.times
[docs] def get_nrecords(self): """ Return the number of records in the file Returns ------- int Number of records in the file. .. deprecated:: 3.8.0 Use :meth:`len` instead. """ warnings.warn( "get_nrecords is deprecated; use len(obj) instead.", DeprecationWarning, ) return len(self)
[docs] def get_residual(self, totim, scaled=False): """ Return an array the size of the model grid containing the flow residual calculated from the budget terms. Residual will not be correct unless all flow terms are written to the budget file. Parameters ---------- totim : float Simulation time for which to calculate the residual. This value must be precise, so it is best to get it from the get_times method. scaled : bool If True, then divide the residual by the total cell inflow Returns ------- residual : np.ndarray The flow residual for the cell of shape (nlay, nrow, ncol) """ nlay = self.nlay nrow = self.nrow ncol = self.ncol residual = np.zeros((nlay, nrow, ncol), dtype=float) if scaled: inflow = np.zeros((nlay, nrow, ncol), dtype=float) select_indices = np.asarray(self.recordarray["totim"] == totim).nonzero()[0] for i in select_indices: text = self.recordarray[i]["text"].decode() if self.verbose: print(f"processing {text}") flow = self.get_record(idx=i, full3D=True) if ncol > 1 and "RIGHT FACE" in text: residual -= flow[:, :, :] residual[:, :, 1:] += flow[:, :, :-1] if scaled: idx = np.asarray(flow < 0.0).nonzero() inflow[idx] -= flow[idx] idx = np.asarray(flow > 0.0).nonzero() l, r, c = idx idx = (l, r, c + 1) inflow[idx] += flow[idx] elif nrow > 1 and "FRONT FACE" in text: residual -= flow[:, :, :] residual[:, 1:, :] += flow[:, :-1, :] if scaled: idx = np.asarray(flow < 0.0).nonzero() inflow[idx] -= flow[idx] idx = np.asarray(flow > 0.0).nonzero() l, r, c = idx idx = (l, r + 1, c) inflow[idx] += flow[idx] elif nlay > 1 and "LOWER FACE" in text: residual -= flow[:, :, :] residual[1:, :, :] += flow[:-1, :, :] if scaled: idx = np.asarray(flow < 0.0).nonzero() inflow[idx] -= flow[idx] idx = np.asarray(flow > 0.0).nonzero() l, r, c = idx idx = (l + 1, r, c) inflow[idx] += flow[idx] else: residual += flow if scaled: idx = np.asarray(flow > 0.0).nonzero() inflow[idx] += flow[idx] if scaled: residual_scaled = np.zeros((nlay, nrow, ncol), dtype=float) idx = inflow > 0.0 residual_scaled[idx] = residual[idx] / inflow[idx] return residual_scaled return residual
[docs] def close(self): """ Close the file handle """ self.file.close()
[docs] def reverse(self, filename: Optional[PathLike] = None): """ Reverse the time order and signs of the currently loaded binary cell budget file. If a file name is not provided or if the provided name is the same as the existing filename, the file will be overwritten and reloaded. Notes ----- While `HeadFile.reverse()` reverses only the temporal order of head data, this method must reverse not only the order but also the sign (direction) of the model's intercell flows. filename : str or PathLike, optional Path of the reversed binary cell budget file. """ filename = ( Path(filename).expanduser().absolute() if filename is not None else self.filename ) # header array formats dt1 = np.dtype( [ ("kstp", np.int32), ("kper", np.int32), ("text", "S16"), ("ndim1", np.int32), ("ndim2", np.int32), ("ndim3", np.int32), ("imeth", np.int32), ("delt", np.float64), ("pertim", np.float64), ("totim", np.float64), ] ) dt2 = np.dtype( [ ("text1id1", "S16"), ("text1id2", "S16"), ("text2id1", "S16"), ("text2id2", "S16"), ] ) nrecords = len(self) target = filename time = ModelTime.from_headers(self.recordarray) time._set_totim_dict() trev = time.reverse() trev._set_totim_dict() nper = time.nper seen = set() def reverse_header(header): """Reverse period, step and time fields in the record header""" nonlocal seen kper = header["kper"] - 1 kstp = header["kstp"] - 1 header = header.copy() header["kper"] = nper - kper header["kstp"] = time.nstp[kper] - kstp kper = header["kper"] - 1 kstp = header["kstp"] - 1 seen.add((kper, kstp)) header["pertim"] = trev._pertim_dict[kper, kstp] header["totim"] = trev._totim_dict[kper, kstp] return header # if rewriting the same file, write # temp file then copy it into place inplace = filename == self.filename if inplace: temp_dir_path = Path(tempfile.gettempdir()) temp_file_path = temp_dir_path / filename.name target = temp_file_path with open(target, "wb") as f: # loop over budget file records in reverse order for idx in range(nrecords - 1, -1, -1): # load header array header = self.recordarray[idx] header = reverse_header(header) # Write main header information to backward budget file h = header[ [ "kstp", "kper", "text", "ncol", "nrow", "nlay", "imeth", "delt", "pertim", "totim", ] ] # Note: much of the code below is based on binary_file_writer.py h = np.array(h, dtype=dt1) h.tofile(f) if header["imeth"] == 6: # Write additional header information to the backward budget file h = header[["modelnam", "paknam", "modelnam2", "paknam2"]] h = np.array(h, dtype=dt2) h.tofile(f) # Load data data = self.get_data(idx)[0] data = np.array(data) # Negate flows data["q"] = -data["q"] # Write ndat (number of floating point columns) colnames = data.dtype.names ndat = len(colnames) - 2 dt = np.dtype([("ndat", np.int32)]) h = np.array([(ndat,)], dtype=dt) h.tofile(f) # Write auxiliary column names naux = ndat - 1 if naux > 0: auxtxt = ["{:16}".format(colname) for colname in colnames[3:]] auxtxt = tuple(auxtxt) dt = np.dtype([(colname, "S16") for colname in colnames[3:]]) h = np.array(auxtxt, dtype=dt) h.tofile(f) # Write nlist nlist = data.shape[0] dt = np.dtype([("nlist", np.int32)]) h = np.array([(nlist,)], dtype=dt) h.tofile(f) elif header["imeth"] == 1: # Load data data = self.get_data(idx)[0] data = np.array(data, dtype=np.float64) # Negate flows data = -data else: raise ValueError("not expecting imeth " + header["imeth"]) # Write data data.tofile(f) # if we rewrote the original file, reinitialize if inplace: move(target, filename) self.__init__(filename, self.precision, self.verbose) # noqa: PLC2801