Source code for

# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""An extensible ASCII table reader and writer.
  Classes to read IPAC table format

:Copyright: Smithsonian Astrophysical Observatory (2011)
:Author: Tom Aldcroft (

import re
from collections import OrderedDict, defaultdict
from textwrap import wrap
from warnings import warn

from astropy.table.pprint import get_auto_format_func
from astropy.utils.exceptions import AstropyUserWarning

from . import basic, core, fixedwidth

class IpacFormatErrorDBMS(Exception):
    def __str__(self):
        return "{}\nSee {}".format(

class IpacFormatError(Exception):
    def __str__(self):
        return "{}\nSee {}".format(

class IpacHeaderSplitter(core.BaseSplitter):
    """Splitter for Ipac Headers.

    This splitter is similar its parent when reading, but supports a
    fixed width format (as required for Ipac table headers) for writing.

    process_line = None
    process_val = None
    delimiter = "|"
    delimiter_pad = ""
    skipinitialspace = False
    comment = r"\s*\\"
    write_comment = r"\\"
    col_starts = None
    col_ends = None

    def join(self, vals, widths):
        pad = self.delimiter_pad or ""
        delimiter = self.delimiter or ""
        padded_delim = pad + delimiter + pad
        bookend_left = delimiter + pad
        bookend_right = pad + delimiter

        vals = [" " * (width - len(val)) + val for val, width in zip(vals, widths)]
        return bookend_left + padded_delim.join(vals) + bookend_right

class IpacHeader(fixedwidth.FixedWidthHeader):
    """IPAC table header"""

    splitter_class = IpacHeaderSplitter

    # Defined ordered list of possible types.  Ordering is needed to
    # distinguish between "d" (double) and "da" (date) as defined by
    # the IPAC standard for abbreviations.  This gets used in get_col_type().
    col_type_list = (
        ("integer", core.IntType),
        ("long", core.IntType),
        ("double", core.FloatType),
        ("float", core.FloatType),
        ("real", core.FloatType),
        ("char", core.StrType),
        ("date", core.StrType),
    definition = "ignore"
    start_line = None

    def process_lines(self, lines):
        """Generator to yield IPAC header lines, i.e. those starting and ending with
        delimiter character (with trailing whitespace stripped)"""
        delim = self.splitter.delimiter
        for line in lines:
            line = line.rstrip()
            if line.startswith(delim) and line.endswith(delim):
                yield line.strip(delim)

    def update_meta(self, lines, meta):
        Extract table-level comments and keywords for IPAC table.  See:

        def process_keyword_value(val):
            Take a string value and convert to float, int or str, and strip quotes
            as needed.
            val = val.strip()
                val = int(val)
            except Exception:
                    val = float(val)
                except Exception:
                    # Strip leading/trailing quote.  The spec says that a matched pair
                    # of quotes is required, but this code will allow a non-quoted value.
                    for quote in ('"', "'"):
                        if val.startswith(quote) and val.endswith(quote):
                            val = val[1:-1]
            return val

        table_meta = meta["table"]
        table_meta["comments"] = []
        table_meta["keywords"] = OrderedDict()
        keywords = table_meta["keywords"]

        # fmt: off
        re_keyword = re.compile(
            r'(?P<name> \w+)'
            r'\s* = (?P<value> .+) $',
        # fmt: on
        for line in lines:
            # Keywords and comments start with "\".  Once the first non-slash
            # line is seen then bail out.
            if not line.startswith("\\"):

            m = re_keyword.match(line)
            if m:
                name ="name")
                val = process_keyword_value("value"))

                # IPAC allows for continuation keywords, e.g.
                # \SQL     = 'WHERE '
                # \SQL     = 'SELECT (25 column names follow in next row.)'
                if name in keywords and isinstance(val, str):
                    prev_val = keywords[name]["value"]
                    if isinstance(prev_val, str):
                        val = prev_val + val

                keywords[name] = {"value": val}
                # Comment is required to start with "\ "
                if line.startswith("\\ "):
                    val = line[2:].strip()
                    if val:

    def get_col_type(self, col):
        for col_type_key, col_type in self.col_type_list:
            if col_type_key.startswith(col.raw_type.lower()):
                return col_type
            raise ValueError(
                f'Unknown data type ""{col.raw_type}"" for column "{}"'

    def get_cols(self, lines):
        Initialize the header Column objects from the table ``lines``.

        Based on the previously set Header attributes find or create the column names.
        Sets ``self.cols`` with the list of Columns.

        lines : list
            List of table lines

        # generator returning valid header lines
        header_lines = self.process_lines(lines)
        header_vals = [vals for vals in self.splitter(header_lines)]
        if len(header_vals) == 0:
            raise ValueError(
                "At least one header line beginning and ending with delimiter required"
        elif len(header_vals) > 4:
            raise ValueError("More than four header lines were found")

        # Generate column definitions
        cols = []
        start = 1
        for i, name in enumerate(header_vals[0]):
            col = core.Column(name=name.strip(" -"))
            col.start = start
            col.end = start + len(name)
            if len(header_vals) > 1:
                col.raw_type = header_vals[1][i].strip(" -")
                col.type = self.get_col_type(col)
            if len(header_vals) > 2:
                col.unit = header_vals[2][i].strip() or None  # Can't strip dashes here
            if len(header_vals) > 3:
                # The IPAC null value corresponds to the io.ascii bad_value.
                # In this case there isn't a fill_value defined, so just put
                # in the minimal entry that is sure to convert properly to the
                # required type.
                # Strip spaces but not dashes (not allowed in NULL row per
                null = header_vals[3][i].strip()
                fillval = "" if issubclass(col.type, core.StrType) else "0"
      , fillval,
            start = col.end + 1

            # Correct column start/end based on definition
            if self.ipac_definition == "right":
                col.start -= 1
            elif self.ipac_definition == "left":
                col.end += 1

        self.names = [ for x in cols]
        self.cols = cols

    def str_vals(self):
        if self.DBMS:
            IpacFormatE = IpacFormatErrorDBMS
            IpacFormatE = IpacFormatError

        namelist = self.colnames
        if self.DBMS:
            countnamelist = defaultdict(int)
            for name in self.colnames:
                countnamelist[name.lower()] += 1
            doublenames = [x for x in countnamelist if countnamelist[x] > 1]
            if doublenames != []:
                raise IpacFormatE(
                    "IPAC DBMS tables are not case sensitive. "
                    f"This causes duplicate column names: {doublenames}"

        for name in namelist:
            m = re.match(r"\w+", name)
            if m.end() != len(name):
                raise IpacFormatE(
                    f"{name} - Only alphanumeric characters and _ "
                    "are allowed in column names."
            if self.DBMS and not (name[0].isalpha() or (name[0] == "_")):
                raise IpacFormatE(f"Column name cannot start with numbers: {name}")
            if self.DBMS:
                if name in ["x", "y", "z", "X", "Y", "Z"]:
                    raise IpacFormatE(
                        f"{name} - x, y, z, X, Y, Z are reserved names and "
                        "cannot be used as column names."
                if len(name) > 16:
                    raise IpacFormatE(
                        f"{name} - Maximum length for column name is 16 characters"
                if len(name) > 40:
                    raise IpacFormatE(
                        f"{name} - Maximum length for column name is 40 characters."

        dtypelist = []
        unitlist = []
        nullist = []
        for col in self.cols:
            col_dtype =
            col_unit =
            col_format =

            if col_dtype.kind in ["i", "u"]:
                if col_dtype.itemsize <= 2:
            elif col_dtype.kind == "f":
                if col_dtype.itemsize <= 4:

            if col_unit is None:
            # This may be incompatible with mixin columns
            null = col.fill_values[core.masked]
                auto_format_func = get_auto_format_func(col)
                format_func =, auto_format_func)
                nullist.append((format_func(col_format, null)).strip())
            except Exception:
                # It is possible that null and the column values have different
                # data types (e.g. number and null = 'null' (i.e. a string).
                # This could cause all kinds of exceptions, so a catch all
                # block is needed here

        return [namelist, dtypelist, unitlist, nullist]

    def write(self, lines, widths):
        """Write header.

        The width of each column is determined in Ipac.write. Writing the header
        must be delayed until that time.
        This function is called from there, once the width information is

        for vals in self.str_vals():
            lines.append(self.splitter.join(vals, widths))
        return lines

class IpacDataSplitter(fixedwidth.FixedWidthSplitter):
    delimiter = " "
    delimiter_pad = ""
    bookend = True

class IpacData(fixedwidth.FixedWidthData):
    """IPAC table data reader"""

    comment = r"[|\\]"
    start_line = 0
    splitter_class = IpacDataSplitter
    fill_values = [(core.masked, "null")]

    def write(self, lines, widths, vals_list):
        """IPAC writer, modified from FixedWidth writer"""
        for vals in vals_list:
            lines.append(self.splitter.join(vals, widths))
        return lines

[docs]class Ipac(basic.Basic): r"""IPAC format table. See: Example:: \\name=value \\ Comment | column1 | column2 | column3 | column4 | column5 | | double | double | int | double | char | | unit | unit | unit | unit | unit | | null | null | null | null | null | 2.0978 29.09056 73765 2.06000 B8IVpMnHg Or:: |-----ra---|----dec---|---sao---|------v---|----sptype--------| 2.09708 29.09056 73765 2.06000 B8IVpMnHg The comments and keywords defined in the header are available via the output table ``meta`` attribute:: >>> import os >>> from import ascii >>> filename = os.path.join(ascii.__path__[0], 'tests/data/ipac.dat') >>> data = >>> print(data.meta['comments']) ['This is an example of a valid comment'] >>> for name, keyword in data.meta['keywords'].items(): ... print(name, keyword['value']) ... intval 1 floatval 2300.0 date Wed Sp 20 09:48:36 1995 key_continue IPAC keywords can continue across lines Note that there are different conventions for characters occurring below the position of the ``|`` symbol in IPAC tables. By default, any character below a ``|`` will be ignored (since this is the current standard), but if you need to read files that assume characters below the ``|`` symbols belong to the column before or after the ``|``, you can specify ``definition='left'`` or ``definition='right'`` respectively when reading the table (the default is ``definition='ignore'``). The following examples demonstrate the different conventions: * ``definition='ignore'``:: | ra | dec | | float | float | 1.2345 6.7890 * ``definition='left'``:: | ra | dec | | float | float | 1.2345 6.7890 * ``definition='right'``:: | ra | dec | | float | float | 1.2345 6.7890 IPAC tables can specify a null value in the header that is shown in place of missing or bad data. On writing, this value defaults to ``null``. To specify a different null value, use the ``fill_values`` option to replace masked values with a string or number of your choice as described in :ref:`astropy:io_ascii_write_parameters`:: >>> from import masked >>> fill = [(masked, 'N/A', 'ra'), (masked, -999, 'sptype')] >>> ascii.write(data, format='ipac', fill_values=fill) \ This is an example of a valid comment ... | ra| dec| sai| v2| sptype| | double| double| long| double| char| | unit| unit| unit| unit| ergs| | N/A| null| null| null| -999| N/A 29.09056 null 2.06 -999 2345678901.0 3456789012.0 456789012 4567890123.0 567890123456789012 When writing a table with a column of integers, the data type is output as ``int`` when the column ``dtype.itemsize`` is less than or equal to 2; otherwise the data type is ``long``. For a column of floating-point values, the data type is ``float`` when ``dtype.itemsize`` is less than or equal to 4; otherwise the data type is ``double``. Parameters ---------- definition : str, optional Specify the convention for characters in the data table that occur directly below the pipe (``|``) symbol in the header column definition: * 'ignore' - Any character beneath a pipe symbol is ignored (default) * 'right' - Character is associated with the column to the right * 'left' - Character is associated with the column to the left DBMS : bool, optional If true, this verifies that written tables adhere (semantically) to the `IPAC/DBMS <>`_ definition of IPAC tables. If 'False' it only checks for the (less strict) `IPAC <>`_ definition. """ _format_name = "ipac" _io_registry_format_aliases = ["ipac"] _io_registry_can_write = True _description = "IPAC format table" data_class = IpacData header_class = IpacHeader def __init__(self, definition="ignore", DBMS=False): super().__init__() # Usually the header is not defined in __init__, but here it need a keyword if definition in ["ignore", "left", "right"]: self.header.ipac_definition = definition else: raise ValueError("definition should be one of ignore/left/right") self.header.DBMS = DBMS
[docs] def write(self, table): """ Write ``table`` as list of strings. Parameters ---------- table : `~astropy.table.Table` Input table data Returns ------- lines : list List of strings corresponding to ASCII table """ # Set a default null value for all columns by adding at the end, which # is the position with the lowest priority. # We have to do it this late, because the fill_value # defined in the class can be overwritten by ui.write, "null")) # Check column names before altering self.header.cols = list(table.columns.values()) self.header.check_column_names(self.names, self.strict_names, self.guessing) core._apply_include_exclude_names( table, self.names, self.include_names, self.exclude_names ) # Check that table has only 1-d columns. self._check_multidim_table(table) # Now use altered columns new_cols = list(table.columns.values()) # link information about the columns to the writer object (i.e. self) self.header.cols = new_cols = new_cols # Write header and data to lines list lines = [] # Write meta information if "comments" in table.meta: for comment in table.meta["comments"]: if len(str(comment)) > 78: warn( "Comment string > 78 characters was automatically wrapped.", AstropyUserWarning, ) for line in wrap( str(comment), 80, initial_indent="\\ ", subsequent_indent="\\ " ): lines.append(line) if "keywords" in table.meta: keydict = table.meta["keywords"] for keyword in keydict: try: val = keydict[keyword]["value"] lines.append(f"\\{keyword.strip()}={val!r}") # meta is not standardized: Catch some common Errors. except TypeError: warn( f"Table metadata keyword {keyword} has been skipped. " "IPAC metadata must be in the form {{'keywords':" "{{'keyword': {{'value': value}} }}", AstropyUserWarning, ) ignored_keys = [ key for key in table.meta if key not in ("keywords", "comments") ] if any(ignored_keys): warn( f"Table metadata keyword(s) {ignored_keys} were not written. " "IPAC metadata must be in the form {{'keywords':" "{{'keyword': {{'value': value}} }}", AstropyUserWarning, ) # Usually, this is done in data.write, but since the header is written # first, we need that here. # get header and data as strings to find width of each column for i, col in enumerate(table.columns.values()): col.headwidth = max(len(vals[i]) for vals in self.header.str_vals()) # keep data_str_vals because they take some time to make data_str_vals = [] col_str_iters = for vals in zip(*col_str_iters): data_str_vals.append(vals) for i, col in enumerate(table.columns.values()): # FIXME: In Python 3.4, use max([], default=0). # See: if data_str_vals: col.width = max(len(vals[i]) for vals in data_str_vals) else: col.width = 0 widths = [max(col.width, col.headwidth) for col in table.columns.values()] # then write table self.header.write(lines, widths), widths, data_str_vals) return lines