Purpose

This Python package can be used to download files from the Cencora (formerly Amerisource) secure file transfer site for ingest into clinical data systems.

Downloads are performed from the web-based secure site located at https://secure.amerisourcebergen.com/. FTP is not supported. (There are many easier ways to automate FTP-based downloads.)

Requirements

• Python 3.10 or newer

Installation

Use pip to install the medberg package.

pip install medberg

Usage

Establishing a connection

Import the SecureSite class from the medberg module.

from medberg import SecureSite

Initialize a connection to the secure site by providing a username and password.

con = SecureSite(username='yourname', password='yourpassword')

Reviewing files

A list of files is automatically downloaded at connection time and stored in the files variable. Files are represented by objects comprising a name, filesize, and upload date.

print(con.files)
# [File(name=340B037AM1234567890330.TXT, filesize=self.filesize='1.3MB', date='03/30/2025'),  ...]

print(con.files[0].name)
# 340B037AM1234567890330.TXT

print(con.files[0].filesize)
# 1.3MB

print(con.files[0].date)
# datetime.datetime(2025, 3, 30, 8, 13, 58)

The library will attempt to automatically extract additional metadata from the filename describing account type (e.g., 340B, GPO, WAC), file specification (e.g., 037, 039), and account number.

print(con.files[0].account_type)
# 340B

print(con.files[0].specification)
# 037AM

print(con.files[0].account_number)
# 123456789

If the metadata is not present in the filename, the corresponding property will simply evaluate to None.

Downloading files

Any individual file can be downloaded using the get() method of the File class. By default, the file contents will be saved to File.contents. Optional parameters can be specified for the save directory (save_dir) and local filename (save_name). If these are included, the file will also be saved to disk. Five attempts will be made to download the file by default. This can be overriden with the max_tries parameter.

con.files[0].get(save_dir='C:\\Users\\yourname\\Downloads\\',
                 save_name='new_filename.txt',
                 max_tries=10)

Files can also be downloaded using the get_file() method of the SecureSite class. In this case, the file to download must be specified in the first parameter as either an instance of the File class or a string containing the filename as it appears on the remote site. This method returns a bytes object instead of the string saved to File.contents.

# Using a File object
file_to_get = con.files[0]
con.get_file(file_to_get)

# Using a string filename
con.get_file('039A_012345678_0101.TXT')

Filtering files

The list of files obtained from the server can be filtered using the match_files() method, which can take any number of arguments in the format file_property=filter_value. For example, to retrieve all files with account number 123456789, you can call match_files(account_number="123456789"). The result will be a list of File objects matching the specified arguments.

con.match_files(account_number="123456789")
# [File(name=340B037AM1234567890330.TXT, filesize=self.filesize='1.3MB', date='03/30/2025'),  ...]

Files can be matched on any attribute. In cases where the file property type differs from the filter value type, the filter value will be converted to the correct type automatically. For example, the account number above was filtered using a string (as account_number is stored in the file class), but it can just as well be filtered using an integer:

con.match_files(account_number=123456789)
# [File(name=340B037AM1234567890330.TXT, filesize=self.filesize='1.3MB', date='03/30/2025'),  ...]

String filter values can contain a wildcard (*) at the beginning or end of the filter. For example, match_files(file_specification="039*") will match "039", "039A", "039AM", etc.

List and tuple filters will cause a match if any one of the inner values matches. Effectively, this acts as a nested OR filter.

Callables can also be passed to allow for more complex filtering. For example, we can get all files from the current month as follows:

from datetime import datetime

current_month = datetime.now().month
current_year = datetime.now().year
con.match_files(date=lambda x: x >= datetime(current_year, current_month, 1))

Multiple filter arguments can be passed together to create a more specific filter.

To get a single file with the most recent upload time that matches a filter or series of filters, use match_latest_file(). This method takes the same arguments as the match_files() method.

Manipulating files

Once files are downloaded, you can perform row-level manipulations using the File.filter_() method. To do this, you must have already downloaded the target file using get(), otherwise this will be performed for you using the default parameters.

Next, a row pattern must be present in File.row_pattern. This is essentially a regex that defines the named capture groups of each line within the file. When the file is downloaded, the library will attempt to match a row pattern based on the parsed specification. If this fails, you must set it manually, e.g.:

file.row_pattern = RowPattern.ICS_039A

Filters are defined as lambda functions based on row properies. Each Row object contains two properties: raw, which is simply a raw string representation of the row from the file, and parts, which contains the parsed elements from the row in a dictionary. Take the following row as an example:

11111111111222222  333333333444444444

When parsed with the ICS_039A row pattern, parts results as the following:

{
    "ndc11": "11111111111",
    "item_id": "222222",
    "price": "333333333",
    "pack_size": "444444444"
}

You could filter in rows that contain an NDC-11 beginning with 11111 with the following lambda:

file.filter_(lambda row: row.parts['ndc11'].startswith("11111"))

If called as a standalone function, filter_() will open the file, filter rows, and save the result on its own. This happens locally inside the File.contents variable. If the file was saved to disk, the file referenced by the path in File.location will be updated also. If multiple applications of filter_() need to be performed, it's recommended to use a with block, which buffers and saves the file at the beginning and end of the block, respectively.

with file as f:
    # Writing to object and disk occurs only after the final filter is applied
    f.filter_(lambda row: row.parts['ndc11'].startswith("11111"))
    f.filter_(lambda row: int(row.parts['price']) / 1000 > 100)

Exporting to a DataFrame

Once a row pattern has been matched, file contents can be exported to a Pandas DataFrame using the File.to_dataframe() method. As with filter_(), if the file has not already been downloaded, this will be performed for you using the default parameters. Each column of the resulting DataFrame corresponds to a key in the row's parts dictionary.

df = file.to_dataframe()
#           ndc11  item_id      price pack_size
# 0  11111111111   222222  333333333 444444444

This method requires the optional pandas dependency, which can be installed with:

pip install medberg[pandas]

Contributing

Pull requests are welcome. Please ensure all code submitted is formatted with Black and tested with pytest. For major changes, please open an issue first to discuss what you would like to change.

When editing the codebase locally, you may install medberg in development mode to use it in REPLs:

pip install -e '.[dev]'

License

This software is licensed under the MIT License.

Disclaimer

This package and its authors are not afiliated, associated, authorized, or endorsed by Cencora, Inc. All names and brands are properties of their respective owners.