Generation and querying of project tables

A project stores soil-property summaries, seismic displacement realizations and their summaries, and yield coefficients associated with admissible displacements. In ShearTable, the fundamental period of the truncated shear-beam model is evaluated from summarized soil properties; a row labelled by a quantile is not a quantile of the period population. DnDraws retains displacements by demand, geometry, material, model, realization and yield coefficient, whereas DnTable contains means and quantiles. kmaxTable summarizes the yield coefficients obtained by inverting those displacement realizations at the admissible displacement levels.

The project contract is a JSON object that identifies input and table paths and supplies the parameters required by the selected steps. Relative paths, including the contract’s own path, are resolved against the project root. The contract directory does not provide a second path origin. Unknown keys are rejected, and the required keys depend on the steps selected. The contract distinguishes geometries and materials, the hazard model and return periods, the number of realizations, displacement-model switches and admissible displacements expressed as percentages of height. The ShearTable step takes an optional seed, without which the contract alone does not make its table reproducible; the displacement step uses fixed seeds for its site and demand combinations.

processProject(), exposed as newmark process, executes selected steps in the order ShearTable, DnTable, kmaxTable. The displacement step reads the shear table, the hazard product MCETable.Rds and each selected site’s UHSRock.Rds. It publishes DnDraws, PGATable and DnTable for that site. The coefficient step reads the stored displacement realizations and peak-acceleration table together with the shear table, then publishes the site’s kmaxTable. Inputs that no selected step produces must already exist when the run is planned. The hazard inputs are read from their declared locations; project processing writes its tables under path.tables.

Site tables reside under <ID.gmdp>/<siteID>/ within that table directory, while global tables reside at its root. Site selection determines which site calculations run. It does not restrict the global tables to those newly calculated sites: the global displacement and coefficient tables are rebuilt from available corresponding tables of the sites discovered under path.uhs, including unselected sites. A discovered site without the corresponding table is omitted from that assembly. A site coefficient table without its associated peak-acceleration table stops coefficient assembly. Consequently, a global result can combine tables from the current run with previously stored site tables.

With dryRun = TRUE, processing reports the selected steps, sites and destinations without calculating, seeding or writing. When ShearTable is not selected, checks for DnTable inspect the existing shear table and required seismic demand, and checks for kmaxTable inspect the shear-table heights. When ShearTable is selected, the dry run cannot examine the table that the step would produce; the actual run performs those downstream checks after publishing it. A successful dry run therefore does not establish that subsequent calculations and publication will succeed.

Publication replaces files individually. Each file replacement is atomic, but a group of tables and the complete run are not atomic and have no rollback. A later failure leaves earlier published files in place. The returned effects record identifies each published file by step, site, product, path, action and row count. After partial publication, the R error has class newmarkPartialPublication and carries that record; the command writes the record to standard output and the failure to standard error, with exit status 1.

listProjectTables(), exposed as newmark listProjectTables, reports existing table files with their row counts, column names, keys, file sizes and modification times. queryProjectTable(), exposed as newmark query, reads one global or site table and selects rows and columns without changing the stored table. A row passes a column filter when it matches any of that filter’s values, and passes the query when it satisfies every column filter. Numeric filters match stored numbers exactly, while text filters preserve stored labels. The returned rows retain their stored order and types, and the requested column order determines the returned columns. For DnTable and kmaxTable, the query adds the column extrapolated, which is not stored: it is 1 when the ky of a row, or its kmax other than 0, lies outside the calibration range that buildKyLimits() gives its model at the mean site peak ground acceleration of the demand, geometry and material, or, for an ensemble, rigid or flexible row, outside the range of any of its models stored for the same site, demand, geometry and material, and 0 otherwise.