Follow-up up on Parsing oscilloscope binary data. I wrote a simple parser for Tektronix's internal file format .isf. I revised my code to the following:

import numpy as np
import os.path

def parse_curve(isf_file):
    Reads one tektronix .isf file and returns a dictionary containing
    all tags as keys. The actual data is stored in the key "data".
    extensions = set([".isf"])
    if os.path.splitext(isf_file)[-1].lower() not in extensions:
        raise ValueError("File type unkown.")

    with open(isf_file, 'rb') as ifile:
        # read header
        header = {}
        while True:
            name = _read_chunk(ifile, " ")
            if name != ":CURVE":
                value = _read_chunk(ifile, ";")

                assert name not in header
                header[name] = value
                # ":CURVE " is the last tag of the header, followed by
                # '#XYYY' with X being the number of bytes of YYY.
                # YYY is the length of the datastream following in bytes.
                value = ifile.read(2)
                y_str = ifile.read(int(value[-1]))
                value += y_str

                # the number of bytes might be present with or without the
                # preceding header ":WFMPRE:"
                nobytes = header.get("BYT_NR",
                                     header.get(":WFMPRE:BYT_NR", "0")
                assert int(y_str) == int(header["NR_PT"]) * int(nobytes)
                header[name] = value
                currentposition = ifile.tell()

        assert header["ENCDG"] == "BINARY"

        # read data as numpy array
        header["data"] = _read_data(ifile, currentposition, header)

    return header

def _read_chunk(headerfile, delimiter):
    Reads one chunk of header data. Based on delimiter, this may be a tag
    (ended by " ") or the value of a tag (ended by ";").
    prior_delimiter = None
    chunk = []
    while True:
        c = headerfile.read(1)
        if c != delimiter:
            if c == '"':
                # switch delimiter to make sure to parse the whole string
                # enclosed in '"'.
                delimiter, prior_delimiter = c, delimiter
        elif prior_delimiter:
            # switch back delimiter
            delimiter, prior_delimiter = prior_delimiter, None
            return "".join(chunk)

def _read_data(bfile, position, header):
    Reads in the binary data as numpy array.
    Apparently, there are only 1d-signals stored in .isf files, so a 1d-array
    is read.
    # determine the datatype from header tags
    datatype = ">" if header["BYT_OR"] == "MSB" else "<"
    if header["BN_FMT"] == "RI":
        datatype += "i"
        datatype += "u"
    # BYT_NR might be present with preceding header ":WFMPRE:BYT_NR"
    nobytes = header.get("BYT_NR",
                         header.get(":WFMPRE:BYT_NR", "")
    datatype += nobytes
    assert len(datatype) >= 3

    data = np.fromfile(bfile, datatype)
    assert data.size == int(header["NR_PT"])

    # calculate true values
    data = data * float(header["YMULT"]) + float(header["YZERO"])

    return data

In the original question, _read_data did a little more, this part was now moved to another module.

One concern I still have, is whether the assert part when reading the ":CURVE" tag is the best solution:

nobytes = header.get("BYT_NR",
                     header.get(":WFMPRE:BYT_NR", "0")
assert int(y_str) == int(header["NR_PT"]) * int(nobytes)

1 Answer 1


In general it looks good to me; only small nits:

You want to move from asserts to throwing proper, informative exceptions - for instance, showing both expected and actual values.

At the end of if name != ":CURVE":, you can just issue a continue instead of an else block - helps with the indentation.

Instead of making chunk a list and then doing a join at the end, you can make it a string and do string appending.

  • 1
    \$\begingroup\$ On the contrary, appending to a list plus one final join is far more efficient and faster, as strings are immutable. See e.g. waymoot.org/home/python_string . I disagree on dropping the else block as well; having 2 blocks makes it clear that only one alternative is followed through. But that may be a style question. \$\endgroup\$ May 20, 2015 at 17:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.