Skip to main content

I/O for ISIS files in Python

Project description

IOISIS - I/O tools for converting ISIS data in Python

This is a Python library with a command line interface intended to access data from ISIS database files and convert among distinct file formats.

The bruma-mst2jsonl command and the bruma module uses a pre-compiled version of Bruma through JPype, which requires the JVM. The iso and mst modules, as well as the mst2jsonl, jsonl2mst, iso2jsonl and jsonl2iso commands don't require Bruma. Bruma only gets downloaded in its first use.

Command Line Interface (CLI)

To use the CLI command, use ioisis or python -m ioisis. Examples:

# Convert file.mst to a JSONL in the standard output stream
ioisis mst2jsonl file.mst

# Convert file.iso to an ASCII file.jsonl
ioisis iso2jsonl --jenc ascii file.iso file.jsonl

# Convert file.jsonl to file.iso where the JSON lines are like
# {"tag": ["field", ...], ...}
ioisis jsonl2iso file.jsonl file.iso

# Convert active and logically deleted records from file.mst
# to file.iso, selecting records and filtering out fields with jq
ioisis mst2jsonl --all file.mst \
| jq -c 'select(.["35"] == ["PRINT"]) | del(.["901"]) | del(.["540"])'
| ioisis jsonl2iso - file.iso

By default, the input and output are the standard streams, but the bruma-mst2jsonl MST input and the jsonl2mst MST output must be a file name, not a pipe/stream. For the former command, the matching XRF will be found based on the file name. For the latter, the control record is created at the end, which makes the random access a requirement.

There are several other options to these commands intended to customize the process, perhaps the most important of these options is the -m/--mode, which regards to the JSONL field format. The valid values for it are:

  • field (default): Use the raw field value string (ignore the subfield parsing options)
  • pairs: Split the field string as an array of [key, value] subfield pairs
  • nest: Split the field string as a {key: value} object

When used together with --no-number, these 3 modes are respectively similar to the -mt1, -mt2 and -mt3 options of isis2json.

Try ioisis --help for more information.


To load ISIS data, you can use the iter_records function of the respective module:

from ioisis import bruma, iso

# For MST files with Bruma, you must use the filename
for record_dict in bruma.iter_records("file.mst"):

# For ISO files, you can either use a file name
# or any file-like object open in "rb" mode
with open("file.iso", "rb") as raw_iso_file:
    for record_dict in iso.iter_records(raw_iso_file):

See also the iter_raw_tl functions and the mst.StructCreator class for more information on how to load data in a more customized way.

One can generate a single ISO record from a dict of data:

>>> from ioisis import iso
>>> iso.dict2bytes({"1": ["testing"], "8": ["it"]})

See also the mst.StructCreator.build_stream method for information on how to create MST files.

By default, the mst module doesn't use/create XRF files. One can create/load XRF data using the struct created by the mst.StructCreator.create_xrf_struct method.

ISO construct containers (lower level data access Python API)

The iso module uses the Construct library, which makes it possible to create a declarative "structure" object that can perform bidirectional building/parsing of bytestrings (instances of bytes) or streams (files open in the "rb" mode) from/to construct containers (dictionaries).

Building and parsing a single record

This low level data access doesn't perform any string encoding/decoding, so every value in the input dictionary used for building some ISO data should be a raw bytestring. Likewise, the parser doesn't decode the encoded strings (tags, fields and metadata), keeping bytestrings in the result.

Here's an example with a record in the "minimal" format expected by the ISO builder. The values are bytestrings, and each directory entry matches its field value based on their index.

>>> lowlevel_dict = {
...     "dir": [{"tag": b"001"}, {"tag": b"555"}],
...     "fields": [b"a", b"test"],
... }

# Build a single ISO record bytestring from a construct.Container/dict
>>> iso_data =
>>> iso_data

# Parse a single ISO record bytestring to a construct.Container
>>> con = iso.DEFAULT_RECORD_STRUCT.parse(iso_data)

# The construct.Container instance inherits from dict.
# The directory and fields are instances of construct.ListContainer,
# a class that inherits from list.
>>> [directory["tag"] for directory in con["dir"]]
[b'001', b'555']
>>> con.fields  # Its items can be accessed as attributes
ListContainer([b'a', b'test'])
>>> len(con.fields) == con.num_fields == 2  # A computed attribute

# This function directly converts that construct.Container object
# to a dictionary of already decoded strings in the the more common
# {tag: [field, ...], ..} format (default ISO encoding is cp1252):
>>> iso.con2dict(con).items()  # It's a defaultdict(list)
dict_items([('1', ['a']), ('555', ['test'])])

Other record fields

Each ISO record is divided in 3 parts:

  • Leader (24 bytes header with metadata)
  • Directory (metadata for each field value, mainly its 3-bytes tag)
  • Fields (the field values themselves as bytestrings)

The leader has:

  • Single character metadata (status, type, coding)
  • Two numeric metadata (indicator_count and identifier_len), which should range only from 0 to 9
  • Free room for "vendor-specific" stuff as bytestrings: custom_2 and custom_3, where the numbers are their size in bytes
  • An entry map, i.e., the size of each field of the directory: len_len, pos_len and custom_len, which should range only from 0 to 9
  • A single byte, reserved, literally reserved for future use
>>> con.len_len, con.pos_len, con.custom_len
(4, 5, 0)

Actually, the reserved is part of the entry map, but it has no specific meaning there, and it doesn't need to be a number. Apart from the entry map and the not included length/address fields, none of these metadata has any meaning when reading the ISO content, and they're all filled with zeros by default (the ASCII zero when they're strings).

>>> con.status, con.type, con.coding, con.indicator_count
(b'0', b'0', b'0', 0)

Length and position fields that are stored in the record (total_len, base_addr, dir.len, dir.pos) are computed in build time and checked on parsing. We don't need to worry about these fields, but we can read them if needed. For example, one directory record (a dictionary) has this:

>>> con.dir[1]
Container(tag=b'555', len=5, pos=2, custom=b'')

As the default dir.custom field has zero length, it's not really useful for most use cases. Given that, we've already seen all the fields there are in the low level ISO representation of a single record.

Tweaking the field lengths

The ISO2709 specification tells us that a directory entry should have exactly 12 bytes, which means that len_len + pos_len + custom_len should be 9. However, that's not an actual restriction for this library, so we don't need to worry about that, as long as the entry map have the correct information.

Let's customize the length to get a smaller ISO with some data in the custom field of the directory, using a 8 bytes directory:

>>> dir8_dict = {
...     "len_len": 1,
...     "pos_len": 3,
...     "custom_len": 1,
...     "dir": [{"tag": b"001", "custom": b"X"}, {"tag": b"555"}],
...     "fields": [b"a", b"test"],
... }
>>> dir8_iso =
>>> dir8_iso
>>> dir8_con = iso.DEFAULT_RECORD_STRUCT.parse(dir8_iso)
>>> dir8_con.dir[0]
Container(tag=b'001', len=2, pos=0, custom=b'X')
>>> dir8_con.dir[1]  # The default is always zero!
Container(tag=b'555', len=5, pos=2, custom=b'0')
>>> dir8_con.len_len, dir8_con.pos_len, dir8_con.custom_len
(1, 3, 1)

What happens if we try to build from a dictionary that doesn't fit with the given sizes?

>>> invalid_dict = {
...     "len_len": 1,
...     "pos_len": 9,
...     "dir": [{"tag": b"555"}],
...     "fields": [b"a string with more than 9 characters"],
... }
Traceback (most recent call last):
construct.core.StreamError: Error in path (building) -> dir -> len
bytes object of wrong length, expected 1, found 2

ISO files, line breaking and delimiters

The ISO files usually have more than a single record. However, these files are created by simply concatenating ISO records. That simple: concatenating two ISO files should result in another valid ISO file with all the records from both.

Although that's not part of the ISO2709 specification, the iso.DEFAULT_RECORD_STRUCT parser/builder object assumes that:

  • All lines of a given record but the last one must have exactly 80 bytes, and a line feed (\x0a) must be included after that;
  • Every line must belong to a single record;
  • The last line of a single record must finish with a \x0a.

That's the behavior of iso.LineSplitRestreamed, which "wraps" internally the record structure to give this "line splitting" behavior, but that can be avoided by setting the line_len to None or zero when creating a custom record struct.

Parsing/building data with meaningful line breaking characters

Suppose we want to store these values:

>>> newline_info_dict = {
...     "dir": [{"tag": b"SIZ"}, {"tag": b"SIZ"}, {"tag": b"SIZ"}],
...     "fields": [b"linux^c\n^s1", b"win^c\r\n^s2", b"mac^c\r^s1"],
... }

That makes sense as an example of an ISO record with three SIZ fields, each with three subfields, where the second subfield is the default newline character of some environment, and the third subfield is its size. Although can build that using the DEFAULT_RECORD_STRUCT (the end of line never gets mixed with the content), we know beforehand that our values have newline characters, and we might want an alternative struct without that "wrapped" line breaking behavior:

>>> breakless_struct = iso.create_record_struct(line_len=0)
>>> newline_info_iso =
>>> newline_info_iso
>>> newline_info_con = breakless_struct.parse(newline_info_iso)
>>> newline_info_simple_dict = dict(iso.con2dict(newline_info_con))
>>> newline_info_simple_dict
{'SIZ': ['linux^c\n^s1', 'win^c\r\n^s2', 'mac^c\r^s1']}
>>> newline_info_iso == iso.dict2bytes(
...     newline_info_simple_dict,
...     record_struct=breakless_struct,
... )

Parsing/building with a custom line breaking and delimiters

The default builder/parser for a single record was created with:

DEFAULT_RECORD_STRUCT = iso.create_record_struct(

We can create a custom object using other values. To use it, we'll pass that object as the record_struct keyword argument when calling the functions.

>>> simple_data = {
...     "OBJ": ["mouse", "keyboard"],
...     "INF": ["old"],
...     "SIZ": ["34"],
... }
>>> custom_struct = iso.create_record_struct(
...     field_terminator=b";",
...     record_terminator=b"@",
...     line_len=20,
...     newline=b"\n",
... )
>>> simple_data_iso = iso.dict2bytes(
...     simple_data,
...     record_struct=custom_struct,
... )
>>> from pprint import pprint
>>> pprint(simple_data_iso.decode("ascii"))
>>> simple_data_con = custom_struct.parse(simple_data_iso)
>>> simple_data == iso.con2dict(simple_data_con)

The calculated sizes don't count the extra line breaking characters:

>>> simple_data_con.total_len, simple_data_con.base_addr
(96, 73)

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

ioisis-0.3.0.tar.gz (38.0 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page