xtgeo 4.14.0__cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

xtgeo/grid3d/_find_gridprop_in_eclrun.py

@@ -0,0 +1,625 @@
+import functools
+import itertools
+import operator
+import pathlib
+import warnings
+from pathlib import Path
+from typing import Dict, List, Literal, Optional, Union
+
+import numpy as np
+import resfo
+
+import xtgeo
+
+from ._ecl_inte_head import InteHead
+from ._ecl_logi_head import LogiHead
+from ._ecl_output_file import Phases
+from ._grdecl_format import match_keyword
+
+sat_keys = ["SOIL", "SGAS", "SWAT"]
+
+
+def filter_lgr(generator):
+    try:
+        while True:
+            entry = next(generator)
+            if entry.read_keyword() == "LGR":
+                warnings.warn(
+                    "Found LGR in ecl run file. "
+                    "LGR's are not directly supported, "
+                    "instead only global values are imported."
+                )
+                while entry.read_keyword() != "ENDLGR":
+                    entry = next(generator)
+            else:
+                yield entry
+    except StopIteration:
+        return
+
+
+# Based on which phases are present some saturation values are given default
+# values, e.g. if phases=Phases.OIL, the saturation of oil ("SOIL") is 1.0 and
+# all other phases are 0.0.
+DEFAULT_SATURATIONS = {
+    Phases.OIL: {
+        "SOIL": 1.0,
+        "SWAT": 0.0,
+        "SGAS": 0.0,
+    },
+    Phases.GAS: {
+        "SOIL": 0.0,
+        "SWAT": 0.0,
+        "SGAS": 1.0,
+    },
+    Phases.WATER: {
+        "SOIL": 0.0,
+        "SWAT": 1.0,
+        "SGAS": 0.0,
+    },
+    Phases.OIL_WATER: {
+        "SGAS": 0.0,
+    },
+    Phases.OIL_GAS: {
+        "SWAT": 0.0,
+    },
+    Phases.GAS_WATER: {
+        "SOIL": 0.0,
+    },
+    Phases.OIL_WATER_GAS: {},
+    Phases.E300_GENERIC: {},
+}
+
+
+def remainder_saturations(saturations):
+    """Infers remainder saturations based on sum(saturations.values()) == 1.
+
+    >>> remainder_saturations({'SWAT': 0.5, 'SGAS': 0.5})
+    {'SOIL': 0.0}
+
+    Args:
+        saturations: Dictionary of phases, such as returned by
+            :meth:`default_saturations()`.
+
+    Returns:
+        dictionary of saturation values that can be inferred.
+    """
+    if all(k in saturations for k in sat_keys):
+        return {}
+    if any(k not in sat_keys for k in saturations):
+        raise ValueError(f"Unknown saturation keys: {list(saturations.keys())}")
+    rest = sum(saturations.values())
+    if len(saturations) == 2 or np.allclose(rest, 1.0):
+        missing = set(sat_keys).difference(set(saturations.keys()))
+        return dict.fromkeys(missing, 1.0 - rest)
+    return {}
+
+
+def peek_headers(generator):
+    """Reads header from a resfo keyword generator without consuming keywords.
+
+    Args:
+        generator: keyword generator such as returned by resfo.lazy_read
+
+    Returns:
+        Tuple of inthead, logihead, and a modified generator which contains all original
+        keywords.
+    """
+
+    def read_headers(generator):
+        intehead_array = None
+        logihead_array = None
+        while intehead_array is None or logihead_array is None:
+            entry = next(generator)
+            kw = entry.read_keyword()
+            if match_keyword(kw, "LOGIHEAD"):
+                logihead_array = entry.read_array()
+            if match_keyword(kw, "INTEHEAD"):
+                intehead_array = entry.read_array()
+
+        intehead = InteHead(intehead_array)
+        logihead = LogiHead.from_file_values(logihead_array, intehead.simulator)
+        return intehead, logihead
+
+    header_generator, generator = itertools.tee(generator)
+    try:
+        intehead, logihead = read_headers(header_generator)
+    except StopIteration as stopit:
+        raise ValueError("Reached end of file without reading headers") from stopit
+    return intehead, logihead, generator
+
+
+def get_fetch_names(name: str) -> List[str]:
+    """Given a gridproperty name, give list of supporting keyword names.
+
+    >>> get_fetch_names('PORO')
+    ['PORO']
+    >>> get_fetch_names('SWAT')
+    ['SOIL', 'SGAS', 'SWAT']
+
+    Args:
+        name: The name of a grid property
+    Returns:
+        List of grid properties that must be fetched from the file.
+
+    """
+    if any(match_keyword(name, saturation_keyword) for saturation_keyword in sat_keys):
+        fetch_names = sat_keys
+    else:
+        fetch_names = [name]
+    return fetch_names
+
+
+def read_values(generator, intehead, names, lengths="all"):
+    """Read the given list of parameter values from the generator.
+
+    Reads the given list of values from the generator. Some saturation
+    values may be inferred from the invariant sum(saturations.values()) == 1.0
+    (see  :meth:`remainder_saturations()`.
+
+    """
+
+    def flatten(lst):
+        return functools.reduce(operator.iconcat, lst, [])
+
+    fetch_names = set(flatten([get_fetch_names(name) for name in names]))
+    defaulted = []
+    if names == "all":
+        values = {}
+    else:
+        values = DEFAULT_SATURATIONS[intehead.phases].copy()
+        defaulted = list(values.keys())
+
+    for entry in generator:
+        if all(name in values for name in fetch_names):
+            break
+        kw = entry.read_keyword()
+        if lengths != "all" and entry.read_length() not in lengths:
+            continue
+        if names == "all":
+            key = kw.rstrip()
+            array = entry.read_array()
+            if np.issubdtype(array.dtype, np.number) or np.issubdtype(
+                array.dtype, bool
+            ):
+                values[key] = entry.read_array()
+        else:
+            matched = [name for name in fetch_names if match_keyword(kw, name)]
+            if len(matched) == 1:
+                if matched[0] in values and matched[0] not in defaulted:
+                    raise ValueError(f"Found duplicate keyword {matched[0]}")
+                values[matched[0]] = entry.read_array()
+            elif len(matched) > 1:
+                # This should not happen if get_fetch_names and
+                # match_keyword work as intended
+                raise ValueError(f"Ambiguous keywords {matched} matched vs {kw}")
+
+    if names == "all":
+        # A more consistent behavior would be to include saturations calculated
+        # with remainder_saturations when names=="all" aswell, however, we do
+        # not include those to keep backwards compatability.
+        # TODO: deprecate this behavior
+        return values
+    values.update(
+        **remainder_saturations({k: values[k] for k in sat_keys if k in values})
+    )
+    return {name: values[name] for name in names if name in values}
+
+
+def check_grid_match(intehead: InteHead, logihead: LogiHead, grid):
+    """Checks that the init/restart headers matches the grid
+
+    Checks that the values given in the headers are compatible with
+    the grid.
+    """
+    dimensions = intehead.num_x, intehead.num_y, intehead.num_z
+
+    if logihead.dual_porosity:
+        dimensions = intehead.num_x, intehead.num_y, intehead.num_z // 2
+
+    if logihead.dual_porosity != grid.dualporo:
+        raise ValueError("Grid dual poro status does not match output file")
+
+    if dimensions != grid.dimensions:
+        raise ValueError(
+            "Grid dimensions do not match dimensions given in output file,"
+            f" {dimensions} vs {grid.dimensions}"
+        )
+
+
+def expand_scalar_values(value, num_cells, dualporo: bool) -> np.ndarray:
+    """Convert from scalar value to filled array of expected size.
+    Args:
+        value: The potentially scalar value
+        num_cells: The number of cells in the grid
+        dualporo: Whether the model has dual porosity
+    Returns:
+        If value is an array, then returns that array, otherwise
+        calls np.full with the shape determined by dualporo status and
+        number of cells.
+
+    """
+    if dualporo:
+        return np.full(fill_value=value, shape=num_cells * 2)
+    return np.full(fill_value=value, shape=num_cells)
+
+
+def pick_dualporo_values(
+    values: np.ndarray, actind: np.ndarray, num_cells: int, fracture: bool
+) -> np.ndarray:
+    """From array of values in an ecl run file, give the fracture or matrix values.
+
+    Args:
+        values: Array of values from an ecl run file.
+        actind: Array of the indecies of active cells.
+        num_cells: Total number of cells.
+        fracture: Whether to give the fracture or matrix values.
+    Returns:
+        Array of either fracture or matrix values from the input values.
+    """
+    active_size = len(actind)
+    indsize = num_cells if len(values) == 2 * num_cells else active_size
+    if fracture:
+        return values[-indsize:]
+    return values[:indsize]
+
+
+def valid_gridprop_lengths(grid):
+    num_cells = np.prod(grid.dimensions)
+    if grid.dualporo:
+        num_fracture = len(grid.get_dualactnum_indices(fracture=True))
+        num_matrix = len(grid.get_dualactnum_indices(fracture=False))
+        return [2 * num_cells, num_fracture + num_matrix]
+
+    num_active = len(grid.get_actnum_indices())
+    return [num_cells, num_active]
+
+
+def match_values_to_active_cells(
+    values,
+    actind,
+    num_cells,
+) -> np.ndarray:
+    """Expands array of ecl run values to be one-to-one with cells.
+
+    In the ecl run file, the values might be only those for active cells.
+    This funtion expands those values to one for each cell with non-active
+    indecies given the xtgeo.UNDEF/xtgeo.UNDEF_INT value.
+
+    Args:
+        values: Array of values from an ecl run file.
+        actind: Array of the indecies of active cells.
+        num_cells: Total number of cells.
+    Returns:
+        Array of input values, but guaranteed num_cells length.
+
+    """
+    if len(values) != len(actind):
+        raise ValueError(
+            f"Unexpected shape of values in init file: {np.asarray(values).shape}, "
+            f"expected to match grid dimensions {num_cells} or "
+            f"number of active cells {len(actind)}"
+        )
+
+    undef = xtgeo.UNDEF_INT if np.issubdtype(values.dtype, np.integer) else xtgeo.UNDEF
+    result = np.full(fill_value=undef, shape=num_cells, dtype=values.dtype)
+    result[actind] = values
+    return result
+
+
+def make_gridprop_values(values, grid, fracture):
+    """Converts values given in init or restart file to one suitable for GridProperty.
+
+    Args:
+    values: The array read from the file
+    grid: The grid from the ecl run
+    fracture: Whether to get the fracture or matrix values
+
+    Returns:
+        Masked array of values indexed by cell
+    """
+    num_cells = np.prod(grid.dimensions)
+    if np.isscalar(values):
+        values = expand_scalar_values(values, num_cells, grid.dualporo)
+
+    if grid.dualporo:
+        actind = grid.get_dualactnum_indices(fracture=fracture, order="F")
+        values = pick_dualporo_values(values, actind, num_cells, fracture)
+    else:
+        actind = grid.get_actnum_indices(order="F")
+
+    if len(values) != num_cells:
+        values = match_values_to_active_cells(values, actind, num_cells)
+
+    values = values.reshape(grid.dimensions, order="F")
+
+    if grid.dualporo:
+        if fracture:
+            values[grid._dualactnum.values == 1] = 0.0
+        else:
+            values[grid._dualactnum.values == 2] = 0.0
+
+    return np.ma.masked_where(grid.get_actnum().values < 1, values)
+
+
+def date_from_intehead(intehead: InteHead) -> Optional[int]:
+    """Returns date format for use in GridProperty name given intehead."""
+    if any(val is None for val in [intehead.day, intehead.month, intehead.year]):
+        return None
+    return intehead.day + intehead.month * 100 + intehead.year * 10000
+
+
+def gridprop_params(values, name, date, grid, fracture):
+    """Make dictionary of GridProperty parameters from imported values."""
+    result = {}
+    result["name"] = name
+    result["date"] = str(date) if date is not None else None
+    result["fracture"] = fracture
+
+    result["ncol"], result["nrow"], result["nlay"] = grid.dimensions
+    result["dualporo"] = grid.dualporo
+    result["dualperm"] = grid.dualperm
+
+    result["values"] = make_gridprop_values(values, grid, fracture)
+
+    if np.issubdtype(result["values"].dtype, np.integer):
+        uniq = np.unique(values).tolist()
+        codes = dict(zip(uniq, uniq))
+        codes = {key: str(val) for key, val in codes.items()}
+        result["codes"] = codes
+        result["values"] = result["values"].astype(np.int32)
+        result["discrete"] = True
+    else:
+        result["codes"] = {}
+        result["values"] = result["values"].astype(np.float64)
+        result["discrete"] = False
+    return result
+
+
+def get_actnum_from_porv(init_filelike, grid):
+    """Override actnum value based on the cell pore volume.
+
+    Args:
+        init_filelike: The init file
+        grid: The grid used by the simulator to produce the init file.
+    Returns:
+        None
+    """
+    generator = filter_lgr(resfo.lazy_read(init_filelike))
+    intehead, logihead, generator = peek_headers(generator)
+
+    if "INTERSECT" not in str(intehead.simulator):
+        # no need to continue if other simulators than INTERSECT
+        return
+
+    check_grid_match(intehead, logihead, grid)
+    porv = read_values(
+        generator, intehead, ["PORV"], lengths=valid_gridprop_lengths(grid)
+    )
+
+    original_active = grid.nactive
+    if porv:
+        if grid.dualporo:
+            num_cells = np.prod(grid.dimensions)
+            actnum_matrix = np.where(
+                porv["PORV"][:num_cells].reshape(grid.dimensions, order="F") > 0.0, 1, 0
+            )
+            actnum_fracture = np.where(
+                porv["PORV"][num_cells:].reshape(grid.dimensions, order="F") > 0.0, 2, 0
+            )
+            grid._dualactnum.values = actnum_matrix + actnum_fracture
+        else:
+            acttmp = grid.get_actnum().copy()
+            acttmp.values = np.where(
+                porv["PORV"].reshape(grid.dimensions, order="F") > 0.0, 1, 0
+            )
+            grid.set_actnum(acttmp)
+
+    new_active = grid.nactive
+    if new_active != original_active:
+        warnings.warn(
+            "The original active cells (ACTNUM) has been overrided by PORV criteria "
+            "from INIT file (purpose: fix simulator INTERSECT issue)",
+            UserWarning,
+        )
+
+
+def find_gridprop_from_init_file(
+    init_filelike,
+    names: Union[List[str], Literal["all"]],
+    grid,
+    fracture: bool = False,
+) -> List[Dict]:
+    """Finds all parameters in a init matching names.
+
+    Note: Does not check that all names are found.
+
+    Args:
+        init_filelike: The init file
+        names: List of property names to be imported. Can also,
+            be set to "all" to import all parameters.
+        grid: The grid used by the simulator to produce the init file.
+        fracture: If a dual porosity module, indicates that the fracture
+            (as apposed to the matrix) grid property should be imported.
+    Returns:
+        List of GridProperty parameters matching the names.
+
+    """
+    init_stream = not isinstance(init_filelike, (str, Path))
+    if init_stream:
+        orig_pos = init_filelike.tell()
+    get_actnum_from_porv(init_filelike, grid)
+    if init_stream:
+        init_filelike.seek(orig_pos, 0)
+    generator = filter_lgr(resfo.lazy_read(init_filelike))
+    intehead, logihead, generator = peek_headers(generator)
+
+    check_grid_match(intehead, logihead, grid)
+
+    date = date_from_intehead(intehead)
+    return [
+        gridprop_params(v, name, date, grid, fracture)
+        for name, v in read_values(
+            generator, intehead, names, lengths=valid_gridprop_lengths(grid)
+        ).items()
+    ]
+
+
+def section_generator(generator):
+    """Sections the generator as delimited by "SEQNUM" keyword.
+
+    Unified restart files will repeat properties in sections, one for each
+    date. The "SEQNUM" keyword indicates a new section. section_generator
+    takes a keyword generator and returns a generator over sections.
+
+    Note: It does so in a lighweight manner so that a section
+    is emptied once the next section is requested, and the original
+    generator is iterated as you progress in the sections.
+    """
+    try:
+        first_seq = next(generator)
+        # for a few (rare?) cases, restarts start with TNAVHEAD, not SEQNUM
+        if match_keyword(first_seq.read_keyword(), "TNAVHEAD"):
+            first_seq = next(generator)
+
+        if not match_keyword(first_seq.read_keyword(), "SEQNUM"):
+            raise ValueError("Restart file did not start with SEQNUM.")
+    except StopIteration:
+        return
+    while True:
+        this_section = itertools.takewhile(
+            lambda x: not match_keyword(x.read_keyword(), "SEQNUM"), generator
+        )
+        yield this_section
+        # empty the section iterator
+        for _ in this_section:
+            pass
+        # peek ahead to see if there are more elements
+        # and stop if there are not
+        try:
+            entry = next(generator)
+            generator = itertools.chain(iter([entry]), generator)
+        except StopIteration:
+            return
+
+
+def find_gridprops_from_restart_file_sections(
+    sections,
+    names: Union[List[str], Literal["all"]],
+    dates: Union[List[int], Literal["all", "last", "first"]],
+    grid,
+    fracture: bool = False,
+) -> List[Dict]:
+    """Finds list of parameters from an sections generator.
+
+    See :meth:`section_generator` for suitable input generator.
+
+    When there are multiple steps/properties matching the given
+    date, the first property matching the date is selected.
+
+    Args:
+        sections: Section generator such as returned by :meth:`section_generator`.
+        names: List of property names to be imported. Can also,
+            be set to "all" to import all parameters.
+        dates: List of xtgeo style dates (e.g. int(19990101)), can also
+            be "all", "last" and "first".
+        grid: The grid used by the simulator to produce the init file.
+        fracture: If a dual porosity module, indicates that the fracture
+            (as apposed to the matrix) grid property should be imported.
+    Returns:
+        List of GridProperty parameters matching the names.
+    """
+    first_date = None
+    last_date = None
+    read_properties = {}
+    for section in sections:
+        intehead, logihead, section = peek_headers(section)
+        check_grid_match(intehead, logihead, grid)
+        date = date_from_intehead(intehead)
+
+        if dates not in ("all", "first", "last") and date not in dates:
+            continue
+
+        section_properties = {
+            (name, date): gridprop_params(v, name, date, grid, fracture)
+            for name, v in read_values(
+                section, intehead, names, lengths=valid_gridprop_lengths(grid)
+            ).items()
+        }
+
+        if dates == "first":
+            if first_date is None:
+                first_date = date
+            elif date != first_date:
+                break
+
+        elif dates == "last" and date != last_date:
+            last_date = date
+            read_properties = {}
+
+        for key in section_properties:
+            if key not in read_properties:
+                read_properties[key] = section_properties[key]
+    return list(read_properties.values())
+
+
+def find_gridprops_from_restart_file(
+    restart_filelike,
+    names: Union[List[str], Literal["all"]],
+    dates: Union[List[int], Literal["all", "first", "last"]],
+    grid,
+    fracture: bool = False,
+    fformat: resfo.Format = resfo.Format.UNFORMATTED,
+) -> List[Dict]:
+    """Finds all parameters in a restart file matching the given names and dates.
+
+    Note: Does not check that all names are found.
+
+    Args:
+        restart_filelike: The restart file.
+        names: List of property names to be imported. Can also,
+            be set to "all" to import all parameters.
+        dates: List of xtgeo style dates (e.g. int(19990101)), can also
+            be "all", "last" and "first".
+        grid: The grid used by the simulator to produce the restart file.
+        fracture: If a dual porosity module, indicates that the fracture
+            (as apposed to the matrix) grid property should be imported.
+    Returns:
+        List of GridProperty parameters matching the names and dates.
+    """
+    close = False
+    if isinstance(restart_filelike, (pathlib.Path, str)):
+        if fformat == resfo.Format.UNFORMATTED:
+            filehandle = open(restart_filelike, "rb")  # noqa: SIM115
+            close = True
+            assume_init_file = (str(restart_filelike)).replace(".UNRST", ".INIT")
+        elif fformat == resfo.Format.FORMATTED:
+            filehandle = open(restart_filelike, "rt", encoding="ascii")  # noqa: SIM115
+            close = True
+            assume_init_file = (str(restart_filelike)).replace(".FUNRST", ".FINIT")
+        else:
+            raise ValueError(f"Unsupported restart file format {fformat}")
+
+        # for INTERSECT, the ACTNUM in the grid is unreliable (bug); hence the ACTNUM is
+        # overriden by reading PORV from the INIT file; note that this will not work
+        # if input is a stream
+        if (Path(assume_init_file)).is_file():
+            get_actnum_from_porv(assume_init_file, grid)
+
+    else:
+        # If restart_filelike is not a filename/path we assume
+        # its a stream
+        filehandle = restart_filelike
+        close = False
+    try:
+        generator = section_generator(filter_lgr(resfo.lazy_read(filehandle)))
+        read_properties = find_gridprops_from_restart_file_sections(
+            generator,
+            names,
+            dates,
+            grid,
+            fracture,
+        )
+    finally:
+        if close:
+            filehandle.close()
+    return read_properties