---

Requirements:

---

• Java 1.7+
• Eclipse OCL 6.0.1+

This repository contains all the necessary scripts and inputs we used to evaluate TD-SB-temPsy tool. The repository contains 2 major steps:

• Data Preprocessing: traces generation and interpolation from Preprocessing folder, based on raw traces and unique signals defined in these traces.
• Tool execution: execution of TD-SB-TemPsy tool from lu.svv.offline folder, given the preprocessed/interpolated traces (.csv format) and temporal properties (.xmi format).

---

Data Preprocessing:

---

Our data preprocessing is implemented in class TracesPreprocessorAndInterpolator.java located under Preprocessing folder. The preprocessing is essentially based on two steps:

---

• Traces Merge:

---

The merge code is written in class TracesMerger.java.

• Function mergeLogsBasedOnSimulationDuration takes in our raw traces RawTraces folder and returns merged traces stored under a generated folder called MergedTraces. The merge consists of merging all the logs from the input raw traces (represented in a .tsv format).
• Function writeToStatisticsFile is used to compute the merge statistics (w.r.t number of entries and the simulation duration per merged log). The latter calls createMergeStatisticsFile and addInformationAfterTracesMerge functions sdefined under TraceStatisticsclass.

---

• Merged Traces Preprocessing:

---

The preprocessing inputs are:

• The merged traces (generated from the previous step), that we store under Preprocessing/MergedTraces folder.
• The xmi properties stored under XmiProperties folder.

Our preprocessing is implemented in TracesPreprocessorAndInterpolator class. Similarly to the traces merger, we first preprocess the traces then we compute the preprocessing statistics.

• Preprocessing: It is based on 5 functions as follows:
  • removeUselessAttributesAndConvertTimestamps: removes the useless columns and converts the date format to microseconds from the raw traces generated under MergedTraces folder. The output traces are stored under PreProcRawData folder.
  • getTracesToInterpolate: for each xmi property from the properties folder XmiProperties, we first retrieve all the unique signals from the property definition, then we loop over the preprocessed traces from PreProcRawData folder. For each of these traces, we store all the records from that trace that concern the unique property signals only to a new trace that we store under CsvLogs folder. It results that each xmi property can come with many generated csv traces from CsvLogs folder. We then generate a match file match.txt that we store under the generated folder Match. The file contains all the properties/traces possible combinations.
  • sort: used to sort the preprocessed traces from CsvLogs folder, based on the timestamp column.
  • applyInterpolation: for each of the preprocessed traces under CsvLogs folder, we apply an interpolation function in order to deal absence of records for missing timestamps. We implemented 2 different interpolations. The choice of the interpolation depends on the required interpolation from each signal definition, from each .xmi property. Our interpolation functions are the following:
    • A linear interpolation.
    • A piecewise-constant interpolation, based on the last-seen signal value.
  • mergeRecordsByTimestamp: the function is used to merge all rows from each log from CsvLogs folder based on similar timestamps.
  • getFinalLogs: used to keep the required columns only in each preprocessed trace. It produces the final traces that we need for our trace diagnostics approach later.
  • evaluateComplexSignals and getFinalTracesWithComplexSignals are functions used evaluate complex signals, represented as a mathematical expression.

---

lu.svv.offline:

---

To run TD-SB-TemPsy, one needs 2 types of inputs:

• properties written in an .xmi format, stored under XmiProperties folder and
• preprocessed/interpolated traces written in a .csv format, stored under CsvLogs folder. Two more folders are needed to run our tool:
• lib: contains the sb-tempsy-report.ocl file that contains the full implementation of the violation causes/ diagnoses semantics definitions.
• models: contains the 4 following ecore models:
  • sbtempsy.ecore: a model generated from our DSL TD-SB-TemPsy-DSL syntax,
  • trace.ecore: a model that defines the data structure of our traces and
  • diagnostics.ecore: a model that contains the full definition of all the violation types supported by TD-SB-TemPsy tool and
  • check.ecore: a model that is used to generate artefacts based on the three models aforementioned. Finally, the main class for running our tool is the following:
• SingleDiagnosticsMain.java: the main class required to report diagnosis for a single violated property over a single trace.

TD-SB-TemPsy tool can be ran using 2 different methods: - from the IDE (e.g., Eclipse environement), the arguments in lines 10 and 11 shall be replaced by the absolute paths of the property and the trace, respectively. - Example: String propertyFile = "./XmiProperties/PropertyExample.xmi"; String traceFile = "./RawTraces/traceExample1.csv"; - as a standalone, using a jar file, where the absolute paths of the property and the trace shall be written in the commandline.

Copyright by University of Luxembourg 2020-2021.

Developed by Chaima Boufaied, chaima.boufaied@uni.lu University of Luxembourg.

Developed by Claudio Menghi, claudio.menghi@uni.lu University of Luxembourg.

Developed by Domenico Bianculli, domenico.bianculli@uni.lu University of Luxembourg.

Developed by Lionel Briand, lionel.briand@uni.lu University of Luxembourg.