Revision history

This chapter describes improvements compared to earlier versions of cutplace.

Version 0.8.8, 2015-11-13

  • Changed development status to “Production/Stable” as cutplace has been processing millions of data rows on a daily base for a couple of months now.
  • Improved validation of Excel dates and times. Fields of type DateTime now check the format specified by the rule and in case it only contains date or time components only extracts those. (API note: internally, Excel dates and times returned by the low level function cutplace.rowio.excel_rows() still use the full YYYY-MM-DD hh:mm:ss format. This change mostly concerns cutplace.validio.Reader and cutplace.fields.DateTimeFieldFormat.validated_value().)
  • Cleaned up CID and data files for documentation, examples and tests ( issue #107 <https://github.com/roskakori/cutplace/issues/107>). There are fewer files now and they have multiple uses. Furthermore examples in the documentation now match the CID’s in the examples folder. Partially this can be attributed to parts of the documentation now including RST files that are updated during the build process.
  • Added Python 3.5 as a supported version.

Version 0.8.7, 2015-07-18

  • Fixed that errors detected by field formats during declaration got suppressed and typically resulted in more confusing errors later.
  • Fixed NameError on platforms without tkinter.
  • Improved documentation:
  • Improved API:
    • Fixed that cutplace.rows() also yielded header rows instead of only data rows.
    • Cleaned up API documentation (typos, links).

Version 0.8.6, 2015-07-14

  • Fixed installation from source distribution by upgrading to Pyscaffold 2.2.1. Earlier versions used versioneer for version numbering and forgot to include versionneer.py in the distribution archive. The current version uses setuptools_scm which only needs a clean dependency instead of additional source code (issue #108).
  • Added command line option --gui to open a graphical user interface for validation (issue #77).
  • Improved error message for broken field declarations in CIDs detected by the field format constructor, which now include the name of the field and the location in the CID.
  • Improved error message when attempting to use a 0 byte as item delimiter in CID’s for delimited data. (Python’s csv module fails with a TypeError because the low level implementation is based on C).
  • Improved API: Cleaned up cutplace.Cid:
    • Changed cutplace.Cid.add_check() to require an cutplace.fields.AbstractCheck as parameter and added cutplace.Cid.add_check_row() to accept a row.
    • Changed cutplace.Cid.add_field_format() to require an cutplace.fields.AbstractFieldFormat as parameter and added cutplace.Cid.add_field_format_row() to accept a row.

Version 0.8.5, 2015-03-09

  • Fixed data format property header, which was ignored (issue #93).
  • Added Constant field format (issue #92).

Version 0.8.4, 2015-03-01

  • Fixed validated writing of header rows by disabling validation of the header.
  • Fixed reading of non ASCII values from ODS under Python 2.
  • Fixed default decimal separator, which now is dot (.) instead of comma (,). Interestingly enough in practice this never really mattered as long as there was no thousands separator (the default), in which case a decimal value using a dot as actual decimal separator simply preserved it and got accepted anyway.
  • Added rule to Decimal fields which allows to specify a range and precision (issue #10, contributed by Patrick Heuberger).
  • Improved documentation: cleaned up section on Using the graphical user interface.
  • Improved API documentation: added a section on Writing data.
  • Improved API: changed validation of length for fixed field values: cutplace.Writer rejects too long values with a FieldValueError and automatically pads too short values with trailing blanks while the low level cutplace.rowio.FixedRowWriter rejects both cases with an AssertionError. Furthermore the length of fixed values is now checked before validating it against the field format rule.

Version 0.8.3, 2015-01-31

  • Added option --until to increase performance by skipping validation of field format and checks after a specified number of rows (issue #86).
  • Fixed reading of Excel error cells.
  • Improved API:
    • Removed shortcuts for exceptions from cutplace. Use the originals in cutplace.errors instead.
    • Added convenience function cutplace.validate() and cutplace.rows() to validate and read data with a single line of source code.
    • Added cutplace.Writer for validated writing of delimited and fixed data (issue #84).
    • Improved API documentation.

Version 0.8.2, 2015-01-19

  • Changed syntax for ranges to prefer ellipsis (...) over colon (:) because it expresses the intended meaning more clearly. The colon is still supported so existing CIDs keep working, but the documentation and examples use the new syntax.
  • Improved error reporting when parsing CIDs. In particular all errors related to the data format include a specific location, and some errors provide more information about the context they occurred in.
  • Cleaned up --help:
    • Removed description of obsolete option --cid-encoding.
    • Cleaned up option groups with only one option.
    • Cleaned up sequence of options which is now sorted alphabetically.
  • Cleaned up notes on Development to reflect changes of 0.8.0.

Version 0.8.1, 2015-01-11

  • Fixed ranges for Integer fields with a length of one digit, which caused a ValueError.
  • Added Python 2 support to universal wheel for distribution.

Version 0.8.0, 2015-01-11

This version is a major rework of the whole code base in order to to fix some long standing bug and migrate it to Python 3.2+ while retaining support for Python 2.6+. A big thank you goes to Patrick Heuberger, Jakob Neuberger and Patrick Prohaska for doing this as a school project for HTL Wiener Neustadt.

In summary, the changes are:

  • A few long standing bugs have finally been fixed, in particular:
    • Fixed that command line client gets stuck on CID in ODS format with syntax error (issue #46)
    • Fixed that delimited format fails when last char of field is escaped (issue #49)
    • Fixed ImportError: No module named xlrd (issue #50)
  • The documentation is now available at http://cutplace.readthedocs.org/en/latest/.
  • Cutplace interface definitions are now abbreviated as CID, replacing the acronym ICD (interface control document). Nevertheless the file format remains the same so existing data descriptions can be used as is.
  • The distribution now uses the wheel format instead of egg. A source distribution is still available as ZIP.

Rarely used functionality that seemed a good idea to have at some time has been removed. If you deem of these features critical, feel free to submit a pull request or to open an issue and request a reimplementation:

  • The cutplace command line options --accept and --reject are gone and all output options related to it. If you still need a filter to build a file that preserves all valid rows and removes rejected ones, a few line of Python code can do the trick:

    from cutplace import Cid, Reader
    cid = Cid('.../some_cid.ods')
    reader = Reader(cid, '.../some_data.csv')
    for row in reader.rows(on_error='continue'):
        # Do something with ``row``.
        pass
    
  • The command line option --listencodings is gone. Instead refer to the standard encodings listed in the Python documentation.

  • The command line option --cid-encoding is gone. If you need non ASCII characters, use ODS format or CSV with UTF-8.

  • The command line option --web (and all related options) to launch a small web server with a validation form is gone. Eventually there is going to be a GUI client, refer to issue #77.

  • The tool cutsniff to build a draft CID is gone as it only takes a few minutes to build a draft anyway. Furthermore, the plain CSV results always needed quite some work to get a more presentable format concerning layout and colors.

The API (see Module Index) has been reworked too and is cleaner and more pythonic now. The project structure applies most of the Simple Rules For Building Great Python Packages. The basic project structure and build process are provided by Pyscaffold.

  • All essential functions can be accessed after a simple import cutplace. The various sub modules are needed only for special requirements.
  • All errors raised by cutplace are collected in cutplace.errors.

Version 0.7.1, 2012-05-20

  • Changed error location of failed row checks to use the first column instead of a number one past the actual number of columns (issue #42).
  • Changed Pattern field format to allow shell patterns instead of only simple DOS patterns (issue #37).
  • Improved cutsniff:
    • Added sniffing of numeric fields (#48).
    • Added first none empty field value as example.
  • Moved project and repository to <https://github.com/roskakori/cutplace> (issue #47).
  • Improved API:
    • Added validating writer, see interface.Writer for more information (issue #45).
    • Added property example for *FieldFormat (issue #41).
  • Cleaned up build and the section on “Jenkins” so that everything works as described even if Jenkins runs as deamon with MacPorts.

Version 0.7.0, 2012-01-09

  • Added command line option --plugins to specify a folder where cutplace looks for plugins declaring additional field formats and checks. For details, see Using your own checks and field formats.
  • Changed interface.validatedRows(..., errors="yield") to yield tools.ErrorInfo in case of error instead of Exception.
  • Reduced memory foot print of CSV reading (Ticket #32). As a side effect, all formats now read and validate in separate threads, which should result in a slight performance improvement on systems with multiple CPU cores.
  • Cleaned up developer reports (Ticket #40). Most of the reports are now built using Jenkins as described in “Jenkins”, the only exception being the profiler report to monitor performance. Also changed build instructions to favor ant over setup.py.
  • Cleaned up API:
    • cutplace and cutsniff have a similar main() that returns an integer exit code without actually calling sys.exit().
  • Cleaned up formatting to conform to PEP8 style.

Version 0.6.8, 2011-07-26

  • Fixed “see also” location in error messages caused by IsUniqueCheck which used the current location as original location.
  • Fixed AttributeError when using the API method AbstractFieldFormat.getFieldValueFor().
  • Fixed ImportError during installation on systems lacking the Python profiler.

Version 0.6.7, 2011-05-24

  • Added option --names to cutsniff to specify field names as comma separated list of names. Without this option, the names found in the last row specified by --head are used. Without this option, fields names will have generated values the user manually will have to change in order to get meaningful names.

Version 0.6.6, 2011-05-18

  • Cleaned up debugging output.

Version 0.6.5, 2011-05-17

  • Added command line option --header to cutsniff to exclude header rows from analysis.
  • Fixed build error in case module coverage was not installed by making coverage a required module again.

Version 0.6.4, 2011-03-19

  • Added cutsniff, a tool to create an ICD by analyzing an existing data file.
  • #21: Fixed automatic detection of Excel format when reading ICDs using the web interface. (Tickte #21).
  • Fixed AttributeError when data format was set to “delimited”.

Version 0.6.3, 2010-10-25

  • Fixed InterfaceControlDocument.checkNames which actually contained the field names. Additionally, checkNames now contains the names in the order they were declared in the ICD. Consequently the checks are performed in this order during validation unlike until now, where the internal hashcode decided the order of checks. (Ticket #35)
  • Improved documentation, in particular:
    • Added more information on writing field format and checks of your own. It still lacks details on how to actually use these in an ICD though. (Ticket #33)
    • Cleaned up introductions of most chapters with the intention to make them easier to comprehend.
  • Changed public instance variables to properties. This allows to mark many of them as read only, and also makes them show up in the API reference. (Ticket #34).

Version 0.6.2, 2010-09-29

  • Added input location for error messages caused by failed checks. (Ticket #26, #27 and #28)
  • Added error message if a field name is a Python keyword such as class or if. This avoids strange error messages if later an IsUnique check refers to such a field. (Ticket #20)
  • Changed style for error messages referring to locations in CSV, ODS and Excel data to R1C1. For example, “R17C23” points to row 15, column 23.
  • Changed internal modules to use “_” as prefix in name. This removes them from the API documentation. Furthermore, module tools has been split into public tools and internal _tools.
  • Changed interface for listeners of validation events:
    • Renamed ValidationListener to BaseValidationListener.
    • Added parameter location to acceptedRow() which is of type tools.InputLocation.
  • Cleaned up API documentation, using reStructured Text as output format and adding a tutorial in chapter Application programmer interface.
  • Cleaned up logging to slightly improve performance.

Version 0.6.1, 2010-04-25

  • Added data format properties “decimal delimiter” (default: ”.”) and “thousands delimiter” (default: none). Fields of type Decimal take them into account. See also: Ticket #24.

  • Added detailed error locations to some errors detected when reading the ICD.

  • Changed choice fields to be case sensitive.

  • Changed choice fields to support values in quotes. That way it is also possible to use escape sequences within values. Values with non ASCII characters (such as umlauts) have to be quotes now. See also: Ticket #25.

  • Renamed module cutplace.range to cutplace.ranges to avoid name clash with the built in Python function range(). In case you have an older version of cutplace installed and plan to import the cutplace Python module using:

    from cutplace import * # ugly, avoid anyway
    

    you will have to manually remove the files cutplace/range.py and cutplace/range.pyc (in case it exists).

  • Added API documentation available from <http://roskakori.github.com/cutplace/api/>.

Version 0.6.0, 2010-03-29

  • Changed license from GPL to LGPL so closed source application can import the cutplace Python module.
  • Fixed validation of empty dates with DateTime fields.
  • Added support for letters, hex numbers and symbolic names in ranges.
  • Added support for letters, escaped characters, hex numbers and symbolic names in item delimiters for data formats.
  • Added auto detection of item delimiters tab (“\t”, ASCII 9) and vertical bar (|). [Josef Wolte]
  • Cleaned up code for field validation.

Version 0.5.8, 2009-10-12

  • Changed Unicode encoding errors to result in the row to be rejected similar to a row with an invalid field instead of a simple message in the console.
  • Changed command line exit code to 1 instead of 0 in case validation errors were found in any data file specified.
  • Changed command line exit code to 4 instead of 0 for errors that could not be handled or reported otherwise (usually hinting at a bug in the code). This case also results in a stack trace to be printed.

Version 0.5.7, 2009-09-07

  • Fixed validation of empty Choice fields that according to the ICD were allowed to be empty but nevertheless were rejected.
  • Fixed a strange error when run using Jython 2.5.0 on certain platforms. The exact message was: TypeError: 'type' object is not iterable.

Version 0.5.6, 2009-08-19

  • Added a short summary at the end of validation. Depending on the result, this can be either for instance eggs.csv: accepted 123 rows or eggs.csv: rejected 7 of 123 rows. 2 final checks failed..
  • Changed default for --log from``info`` to warning.
  • Improved confusing error message when a field value is rejected because of improper length.
  • Fixed ImportError when run using Jython 2.5, which does not support the Python standard module webbrowser. Attempting to use --browser will result in an error message nevertheless.

Version 0.5.5, 2009-07-26

  • Added summary to validation results shown by web interface.
  • Fixed validation of Excel data using the web interface.
  • Cleaned up reporting of errors not related to validation via web interface. The resulting web page now is less cluttered and the HTTP result is a consistent 40x error.

Version 0.5.4, 2009-07-21

  • Fixed --split which did not actually write any files. (Ticket #19)
  • Fixed encoding error when reading data from Excel files that used cell formats of type data, error or time.
  • Fixed validation of Decimal fields, which resulted in a NotImplementedError.
  • Fixed internal handling of ranges with a default, which resulted in a NameError.

Version 0.5.3, 2009-07-18

  • Added command line option --split to store accepted and rejected data in two separated files. See also: ticket #17.
  • Fixed handling of non ASCII data, which did not work properly with all formats. Now cutplace consistently uses Unicode strings to internally represent data items. See also: ticket #18.
  • Improved error messages and removed stack trace in cases where it does not add anything of value such as for I/O errors.
  • Changed development status from alpha to beta.

Version 0.5.2, 2009-06-11

  • Fixed missing setup script.

Version 0.5.1, 2009-06-11

  • Added support for ICDs in Excel and ODS format for built in web server.
  • Changed representation of integer number read from Excel data: instead of for example “123.0” this now renders as “123”.
  • Improved memory usage for data and ICDs in ODS format.
  • Fixed reading of ICDs in Excel and ODS format.
  • Fixed TypeError when the CSV delimiters specified in the ICD were encoded in Unicode.

Version 0.5.0, 2009-06-02

  • Fixed handling of Excel numbers, dates and times. Refer to the section on Excel data format for details.
  • Changed order for field format (again): It now is name/example/empty/length/type/rule instead of name/example/empty/type/length/rule.
  • Changed optional items for field format: now the field name is the only thing required. If no type is specified, “Text” is used.
  • Added a proper tutorial that starts with a very simple ICD and improves it step by step. The old tutorial presented one huge ICD and attempted to explain everything in it, which could easily overwhelm the reader.
  • Migrated documentation from DocBook to RestructuredText.
  • Improved build and installation process (setup.py).

Version 0.4.4, 2009-05-23

  • Fixed checks when validating more than one data file from the command line. Until now the checks did preserve internal state information needed to perform the check. For instance, IsUnique check remembered the keys of all rows read so far. So when a data file contained a row with a key that already showed up in an earlier data file, the check failed. To prevent this from happening, validate() now resets all checks. See also: Ticket #9.
  • Fixed detection of characters outside of the “Allowed characters” range. Apparently this never worked until now.
  • Fixed handling of empty choices consisting only of white space.
  • Fixed detection of fixed fields without length.
  • Fixed handling of white space in field items of fixed length data.
  • Added plenty of test cases and consequently performed a couple of minor fixes, improvements and clean ups.

Version 0.4.3, 2009-05-18

  • Fixed auto detection of delimiters in a CSV file, which got broken when switching to Python’s built in CSV reader with version 0.3.1. See also: Ticket #8.

Version 0.4.2, 2009-05-17

  • Added validation for data format property “Allowed characters”, which can be used with all data formats.
  • Added data format property “Header” to specify the number of header rows that should be skipped without validation. This property can be used with all data formats.
  • Added data format property “Sheet” to specify the number of the sheet to validate in spreadsheet data formats (Excel and ODS).
  • Added complex ranges that consist of several sub ranges separated by a comma (,). For example: “10:20, 30:40” means that a value must be between 10 and 20 or 30 and 40.
  • Moved forums to http://apps.sourceforge.net/phpbb/cutplace/.
  • Moved project site and issue tracker to http://apps.sourceforge.net/trac/cutplace/.
  • Fixed handling of data rows with too few or too many items.
  • Cleaned up error handling and error messages.

Version 0.4.1, 2009-05-10

  • Added support for Excel and ODS data formats.

Version 0.4.0, 2009-05-06

  • Added support for ICDs stored in Excel format. In order for this to work, the xlrd Python package needs to be installed. It is available from http://pypi.python.org/pypi/xlrd.
  • Changed ICD format: Inserted a new column after the field name and before the field type that can contain an optional example value. This enables readers to quickly grasp the meaning of a field by taking a glimpse at the first few columns instead of having to “decipher” the field type and rule.

Version 0.3.1, 2009-05-03

  • Added proper error messages for several possible error the user might make when writing an ICD. So far these errors resulted into confusing messages about failed assertions, attempted NoneType accesses and the like.

  • Added requirement that field names in the ICD only use ASCII letters, digits and underscore (_). This is necessary to prevent Python errors in checks that refer to field values using Python variables, such as DistinctCount and IsUnique.

  • Changed CSV parser to use Python’s built in one. This works around the following issues:

    • Improved performance when working with CSV data (about 4 times faster).
    • Error when reading valid CSV data that contained nothing but a single item separator.

    However, it also introduces new issues:

    • Increased memory usage when working with CSV data because csv.reader does not fit well with the AbstractParser class. Currently the whole file is read into memory.
    • Lack of any error detection in the CSV structure. For example, unclosed quotes at the end or inconsistent line feeds do not raise any errors.
  • On the long run, cutplace will need its own CSV parser. If only this would not be so boring to code...

  • Improved error messages for broken field names and types in the ICD.

Version 0.3.0, 2009-04-28

  • Fixed error messages in case field name or type was missing in ICD.

  • Fixed handling of percent sign (%) in DateTime field format.

  • Changed syntax to specify ranges like field lengths or rules for Integer fields formats. Use ”:” instead of ”...”.

    There are basically two reasons for this change: Firstly, this looks more Python-like and thus more consistent with other parts of the ICD like the “Checks” section which also uses Python syntax in various places. Secondly, this avoids issues with Excel which under certain circumstances changes the 3 characters in ”...” to a single character ellipsis. Using ”:” still is not without issues though: if you use a spreadsheet application to author ICDs, most of them think of a value like “1:60” (which could for example specify a field length between 1 and 60 characters) to refer to a time of 1 hour and 60 minutes. To avoid any confusion, disable the cell format auto detection of the spreadsheet application by changing all cells to contain “Text”.

Version 0.2.2, 2009-04-07

  • Added support to use data encodings other than ASCII by specifying them in the data format section of the ICD using the encoding property.
  • Added support for fixed data format.
  • Added command line option --browse to be used together with --web in order to open the validation page in the web browser.
  • Added command line option --icd-encoding to specify the character encoding to be used with ICDs in CSV format.

Version 0.2.1, 2009-03-29

  • Added support for ICDs in ODS format for command line client.
  • Added cutplace.exe for Windows, which will be generated during installation.
  • Added automatic installation of setuptools when you try to build cutplace using the Subversion repository. This feature is provided by ez_setup.py, which is available from the setuptools site.
  • Fixed cutplace script, which did exit with an ExitQuietlyOptionError for options that just showed some information and exited (such as --help).

Version 0.2.0, 2009-03-27

  • Added option --web and --port to launch web server providing a simple graphical user interface for validation.
  • Changed --listencodings to --list-encodings.

Version 0.1.2, 2009-03-22

  • Added DistinctCount check.
  • Added IsUnique check.
  • Added command line option --trace.
  • Added support to validate an ICD when no data are specified in the command line.
  • Cleaned up error messages.

Version 0.1.1, 2009-03-17

  • Initial release.