Suppress warnings

Fixed

Pandas recently started to complain UserWarning: pandas only support SQLAlchemy connectable(engine/connection) ordatabase string URI or sqlite3 DBAPI2 connectionother DBAPI2 objects are not tested, please consider using SQLAlchemy. We suppress the warning since the connections works as they should. No need to add SQLAlchemy as dependency.

Suppress warning about unverified HTTPS requests if users explicitly sets verifySSL=False.

Testing has been somewhat improved. Warnings are now treated as errors to encourage early fix.

Linux and MacOS

The biggest change in this major release is that Tagreader can be installed for Linux and MacOS platforms in addition to Windows. ODBC requests are only available for Windows due to proprietary drivers from AspenTech and OSISoft, but the Web APIs should in any case be the preferred connection methods for all platforms.

Added

• Should work for Linux and MacOS platforms in addition to Windows
• Python 3.10 supported
• Time can be specified using datetime.datetime objects.

Fixed

• Tables will be installed for all systems and Python versions except MacOS with M1 (arm64) CPUs.
• Slightly improved error handling for requests to Web APIs

Handle invalid JSON from AspenOne

Fixed

AspenOne sometimes returns the invalid value nan in its JSON response. When that happens, tagreader will now pre-process the JSON string to replace any occurrence of the value nan or -nan with a valid NaN and then try to parse again.

Fix PI WebAPI bug

Fixed

• Would previously crash for raw reads with PI Web API if there is no data. Now returns empty dataframe.

Deprecated

• Python 3.6 is no longer supported. Tagreader may currently still work for Python 3.6, but any issues related to 3.6 will not be fixed, and code may be introduced that is not compatible with 3.6.

Add support for Python 3.9

Tagreader can now be installed for Python 3.9. The 'tables' package will not be installed as dependency since it still lacks wheels for Python 3.9 in Windows. Tagreader will still function as normal, except that cache will be disabled.

Added

• Support for Python 3.9

Deprecated

• Support for Python 3.6 will be removed within a few months.
• Support for Python 3.7 may be removed within a few months

Fix searches containing space for aspenone

Fixes

• Search queries containing spaces now return results for aspenone (fixes #76)

Add SSL certs on module import

Starting with this release it should no longer be necessary for users from Equinor to manually run add_statoil_root_certificate() before using the Web APIs. Upon importing the module, the function will be silently called if the running environment is detected to be an Equinor host. There may be special cases where this does not happen. In those cases the function can be called manually, but please also raise an issue so it can be fixed.

Added

• Add root certificate on module import (#64)

Fetch status information

This release enables fetching of status/quality information for values for all four IMS types by specifying get_status = True to read(). This feature should be considered somewhat experimental, and the implementation may be changed at a later time. Please be aware that until further notice whenever get_status = True is specified, the cache is silently bypassed. Please see the manual for details.

Added

• Add optional status output for Aspen ODBC, AspenOne, PI ODBC and PI Web API

Fix invalid data from Aspen Web API

Fixed

Aspen Web API will sometimes return values that are not a proper number, e.g. "1.#QNAN". This was stored as a string, which led to the dataframe being of type object. Fixed by converting all non-valid numbers to proper NaN values.

Fix max number of points for Aspentech Web API

Fixed

This release fixes a bug when requesting more than 10 000 interpolated datapoints using Aspen Web API. The resulting dataframe column would be of type object instead of numeric, which causes problems at least when trying to store results to hdf files. Fix was to enforce the max number of data points returned. This max limit was also increased to 100 000, which is the default max reported by the server.