New loading cache implementation

Switched the loading cache implementation from Guava to Caffeine. Caffeine showed roughly a 20% performance increase in our heavily threaded integration tests.

Caffeine requires Java 8 so that will become the new minimum JDK requirement. In addition, since the loading cache was the main thing being used from Guava that dependency was removed from the library.

Encoding fix

The code which updated the schemas and tables from SEER*API was not handling encoding correctly. This caused issues with certain characters being converted incorrectly in some table notes. The code was corrected and the tables were also updated to fix the bad characters.

TNM v1.1

TNM v1.1 replaces TNM v1.0 in the library.

Changes to version 1.1 include:

• Added schema notes regarding differences between UICC and AJCC chapter definitions and describing SEER*RSA rules.
• Added schema notes related to Summary Stage 2000, including links to manual
• Added footnotes defining the difference among UICC, AJCC and SEER*RSA related to T, N and M
• Added notes defining the difference among UICC, AJCC and SEER*RSA related to directly assigned Stage Group
• Added registrar notes that 88 should be assigned for DCOs to T, N and M
• Corrected typographical errors in various parts of the system. Of import:
  • Liver, schema Note 1: changed to state that C221 with 8170-8175 is NOT TNM staged
  • Thyroid, Directly assigned stage groups: added T0, N1b, M0 -> IVA for Pap>=45 and Med
  • MelanomaConjunctiva, Path T, T1b: description was incorrect, the T1a description had been accidentally copied here
• Corrected Stage group calculation algorithm errors.
  • CorpusSarcoma, TNM Stage Calculation tables and Assigned Stage Group lookups: Changed T3, N1, M0 to map to IIIC
  • Testis, TNM Path Stage Calculation tables and Assigned Stage Group lookups: Changed Tis, (N0, NX), M0, (S0, SX) to map to 0. Only Tis, N0, M0, S0 had gone to 0
  • Combined T General and No TIS tables: where neo-adjuvant treatment was not given, added the combinations of Clin T (1C, 1D) with Path T (1A, 1B)
  • Combined T for Breast: where neo-adjuvant treatment was not given, added the combinations of Clin T (1MI, 1A-1C) with Path T (4, 4A-4D)
  • Combined N General: where neo-adjuvant treatment was given, added the combinations of Clin N (3A-3C) with Path N4
• Thyroid: Adjusted the algorithms, tables and notes to handle Anaplastic tumors with Clin T lower than T4. Clinical stage is calculated using Papillary and Follicular tables for such cases.
• LymphomaOcularAdnexa: Adjusted the algorithms to remove steps related to T, N and M since these are not defined in UICC

Note that the library only contains the latest version of TNM v1.x. If you want to use TNM v1.0 you will need to a prior version of the library.

TNM v1.0

TNM v1.0 is now available.

The TNM staging algorithm contains the data items and descriptions necessary to collect T, N and M stage components, as well as other data items required to calculate Clinical and Pathologic Stage Groups. In addition, this algorithm calculates a combined stage group. It also contains the data item definitions for the associated Predictive and Prognostic Factors on a schema specific basis. Staging information is collected for site and histology combinations that are not defined in TNM in order to support SEER Summary Stage calculations (these calculations will not be implemented for diagnosis year 2016).

The initial release includes Union for International Cancer Control (UICC) TNM Classification of Malignant Tumor 7th edition.

Bugfix release

This release fixes a parsing error for table fields that end in a "-". When parsing cells, values are considered ranges if the length of the data one the left and right of the dash are the same. However the code was not handling values with a dash as the last character. This did not affect the CS algorithm but did affect the upcoming TNM algorithm

Bugfix release

This release fixes a minor issue with table parsing. Cell values with a trailing empty value were not being processing correctly. Values that looked like ,1,2,3 were parsing properly in 4 distinct values, while 1,2,3, was incorrectly parsing into only 3 values. The parsing was fixed to properly handle both cases.

Java 7

This is the first release which requires Java 7.

• There was an exception being thrown in getInvolvedTables when a table had no rows. There was no reason for this and now it is handled properly.

Change how missing context keys are matched

The behavior of missing context keys and how they match tables INPUTs was changed. For a complete description, see #12.

Bugfix for schema lookup staging outputs

While the stage API correctly dealt with the outputs defined on the schema level, the lookupSchema were not.

There were 2 issues corrected in this release:

1. Values calculated in staging but not in the output list were being listed as outputs in lookupSchema. They do not get returned in stage.
2. Outputs that have defaults values (like schema_number in CS) were not being included in the list of outputs returned by lookupSchema.

Minor bugfix

There was an problem with the calculation of inputs to individual table paths when the library was run under Java 8. There were additional inputs listed that should have been ignored. The releases fixes the issues and updates the CS 02.05.50 schema files to incorporate the fix.