qa-dkrz: CORDEX_qa.conf

#
# Configuration file for the Quality Control of CORDEX
# ========================================================
# Note: Default values in []
# Note: may contain deprecated or even obsolete options.

# KEY WORDS controlling the application in the quality-control-run.
# SYNTAX: key[[+][=]value[,value[,value]]
# Detailed explanation of syntax rules at the end of this file.

# Section: Customisation
# ----------------------

# Path to the root of directory tree(s) with netCDF data files.
# Note: the contained sub-directory tree(s) are appended to QA_RESULTS.
# Note: this option is mandatory.
PROJECT_DATA=/home/vagrant/drs

# Path to QA results and logs.
# Note that results will be inegrated into the path to QA-DKRZ by default
# (not recommended).
QA_RESULTS=/tmp/QA_results/zequi

# Selection of paths and variables (RegExp of the 'expr' command,
# i.e. full specification from the beginning of the word).
# Paths and variables are given by '[path1[,path2,..]][=][var1[,var2,..]]'.
# 1) The '=' character is required when both path and variable items are specified.
#    It is also required for paths without any '/'.
# 2) Omission of path components selects variables in all sub-paths
#    cointained in PROJECT_DATA.
# 3) Omission of variable component(s) selects all variables in the path(s).
# 4) By default, all files in PROJECT_DATA (and sub-trees) are selected.
# Full syntax rules and examples  at the end of this file.
# SELECT path=var
# SELECT AFR-44/CLMcom/MPI-ESM-LR/historical/r1i1p1/CCLM4-8-17

# Locking of variables (takes precedence over SELECT).
# See SELECT for a description.
# LOCK .*/Z=   # [disabled]

# Purpose: If the QA package is installed on a file system
# accessed by machines of different architectures and if more than one
# is dedicated to run the QA, then these
# require different names for the directories
# of residing executables. By default: your-path/QA-DKRZ/bin . The
# bin-directories must have been specified during the installation process.
# If the path is relative, the directory has to be a sub-dir of QA-DKRZ.
# If directories are prefixed by the hostname(s) of computer(s) separated
# by ':' from the directory, then multiple directories may be specified
# as comma-separated-list. The name(s) of the computer(s) must be
# available in the list of hosts, i.e. QA_EXEC_HOSTS. A name without a prefix
# is dedicated for the host on which qcManager is running (but could also
# have a prefix).
# You have to compile all executables into all bin-directories specified.
# QA_BIN=bin  # surge:bin,another-host:bin2 #[bin]

# This option forces the QA to check only meta-data on the basis of the
# standard table and Controlled Vocabulary. In case of CORDEX:
# CORDEX_variables_requirement and CORDEX_archive_design.
# There is neither a check on time values nor on data (even the existence of
# data is not tested). No project table will be generated or consulted.
# USE_STRICT


# Section: Files and Tables
# -------------------------

# This option controls whether the second time-stamp in the filename
# expresses the sharp end (deadline) of the period or whether the
# most frequent componente of the period in the stamp has to be extended to the end,
# e.g. 2000-03 <=> 2000-03-31 for extended.
# The difference is explained best by examples.
# 1) Sharp:     2000 - 2001 <=> 2000-01-01T00:00:00 - 2001-01-01T00:00:00
# 2) Extended:  2000 - 2001 <=> 2000-01-01T00:00:00 - 2001-12-31T24:00:00
# If there is a mix for the components of the stamps, then use this
# Syntax: 'Yx-Mx-Dx-hx-mx-sx' with x=s or x=e for sharp and extended
# respectively. APPLY_MAXIMUM_DATE_RANGE without any assignment enables 'extended' for
# the date and sharp for the time component.
  APPLY_MAXIMUM_DATE_RANGE

# CF Conventions used if not specified in the global attributes of the file.
# CF=1.4 # [global attribute]

  CF_CHECK_LIST=CF_check-list.conf
  CF_FOLLOW_RECOMMENDATIONS # [f]

# http://cfconventions.org
  CF_AREA_TYPES=CF_area-type-table.txt
  CF_STD_REGION_NAMES=CF_standardized-region-names.txt
  CF_STANDARD_NAMES=cf-standard-name-table.xml

# User-defined handling of exceptions, i.e. how to proceed with
# notification of QA check results. Explanations in file
# QA-DKRZ/projects/CORDEX/CORDEX_check-list.conf.
# Keys: Flag, L[1-4], D, PT, ST, variable names. Please note that
# any misspelling is considered the name of a variable and would
# throw a flag in most cases.
  CF_NOTE={plev,height,CF_5h,D}

# Apply a unique annotation handling to all entries of the check-list.
# CF_NOTE_ALWAYS=L1,EM

# Limit maximum annotation level to all entries of the check-list.
# CF_NOTE_LEVEL_LIMIT=L1

# The original directory tree structure of the source data is preserved,
# i.e. a symbolic link in the data causes making a corresponding symbolic link
# in the directory tree of the QA results, too. Enable this option if
# symbolic links are to be dereferenced, i.e. genuine data are to
# replace the symbolic links in the QA result tree.
# DEREFERENCE_SYM_LINKS # [disabled]

# If distributed computing is also involved with distributed file system,
# then enable this option. Bi-directional communication and data
# transfer is done by ssh and scp, respectively. IMPORTANT NOTE: this is
# not installed and not projected for CMIP5 or CORDEX.
# But, the QA may have anyway distributed computing with a shared file system.
# DISTRIBUTED_FS # [disabled]

# The directory in the path at which indices of LOG_PATH_INDEX start.
# Note that the first occurrence from left is taken. If there
# are more than a single instance with the same name, and the right most is used.
  DRS_PATH_BASE=CORDEX

# The position (zero-based index) of the frequency within the filename
  FILE_NAME_FREQ_INDEX=7

# The separator for delimiting the filename components
  FILE_NAME_SEP=_

  # The position (zero-based index) of the variable within the filename
  FILE_NAME_VAR_INDEX=0

# By default, paths to data are found ignoring hidden directories.
# HIDDEN_DIRECTORY_DESCENT

# Skip broken links in the directory tree of data,
# i.e. files pointing to a non-existing target.
# By default, a warning is issued.
# IGNORE_BROKEN_LINKS

# Occurrences rising a notification of level L>2 cuase the
# creation of a qc_lock_<filename>.txt file, which will prevent
# any further checking of files of the given variable.
# This option ignores any blocking of following temporal subsets.
# Note: In case that a blocking file was found
# but no qc_<filename>.nc file, then the blocking holds anyway.
# IGNORE_LOCK_FILES

# The purpose of an experiment-like name is just to tag a larger set of various
# checks with a name that corresponds to a certain volume of netCDF files.
# There are several ways to define LOG_NAME: explicite string, respresenting
# a path to a file ensemble, a pattern of data file names, and
# 'undefined-scope' by default (precedence as here).

# Explicit experiment-like name.
# LOG_FNAME=string  # [disabled]

# Based on DRS_PATH_BASE, LOG_NAME is constructed by path components whose indices
# are provided by LOG_PATH_INDEX, a comma-separated sequence with index==1 following
# directly DRS_PATH_BASE.
# Note that the example provides index in [], which are not part of the name.
# For instance, the sequence 1,4,2,3,5 applied to the path
# /base[0]/AFR-44[1]/SMHI[2]/CCCma-CanESM2[3]/historical[4]/r1i1p1[5]/SMHI-RCA4[6]/v1[7]/day[8]/clh[9]
# results in AFR-44_historical_SMHI_CCCma-CanESM2_r1i1p1.
# Note: if any versioning is done in the filename but not in the path,
# then LOG_FNAME_PATTERN should be preferred.
# If LOG_PATH_INDEX remains disabled, then consistency checks are disabled, too.
  LOG_PATH_INDEX=2,3,4,7,5,8,11 # [disabled] note: 11 for versioning if available

# LOG_FILE_INDEX analogously to LOG_PATH_INDEX; the first filename component,
# delimited by LOG_FILE_SEPARATOR, with index=0.
# Note that this method is raised when no DRS_PATH_BASE could be found.
  LOG_FILE_INDEX=1,2,5,3

# Delimiter for filename decomposition; '_' by default.
# LOG_FILE_SEPARATOR='_'

# The separator for delimiting the filename components
  PATH_NAME_SEP=/

# When the name of the PROJECT_TABLE is generated automatically, it will be
# prefixed by 'pt_' by default.
# PROJECT_TABLE_PREFIX=pt

# Purpose: automatic determination of the project table name.
# Note that meta-data of each file will be checked against the consistency table.
# Usage and example as for LOG_PATH_INDEX.
  PT_PATH_INDEX=2,3,4,7,8,11

# This table is telling the QA-program about notification codings, discarding
# certain test, issuing email, and grade of severeness of findings.
# An automatically updated version by default.
  QA_CHECK_LIST=CORDEX_check-list.conf

# The QA ignores empty or non-project data files by default.
# If it is clear that an empty or alien file should never be present in
# in the data directory tree, then issuing messages in the log-files can be enabled.
  QUERY_NON_NC_FILE  # [ disabled]
  QUERY_EMPTY_DIR  # [ disabled]
  QUERY_EMPTY_FILE  # [ disabled]
  QUERY_ONLY_NC  # [ disabled; disables other file queries]

# By default, the QA examines the begin and end dates specified
# by the time records of all sub-temporal files and finds
# by synchronisation the right file for proceeding
# (or starting) a check. The purpose of this option:
# a) check the times in the filenames as well as in the files and
#    search for overlapping periods in the set of sub-temporal files
#    of a given variable.
# b) check the modification times of the sub-temporal files of a given
#    variable. Issue a warning, if the modification time of the set of
#    the sorted files according to a) is not ascending.
#    This is disabled by assigning 'no_mod'.
# c) ignore filename embedded times by assigning 'mixed' (csl accepted)
# SYNC_FILE_AMBIGUITY_CHECK  # [disabled]
# SYNC_FILE_AMBIGUITY_CHECK=no_mod,mixed # [disabled]

# Specification of requirements.
# An automatically updated version by default.
  TABLE_DRS_CV=CORDEX_DRS_CV.csv # CORDEX_archive_design.csv

# List of available GCM model names from
# http://cordex.dmi.dk/joomla/images/CORDEX/GCMModelName.txt
  TABLE_GCM_NAME=CORDEX_GCMModelName.txt

# List of available RCM model names from
# http://cordex.dmi.dk/joomla/images/CORDEX/RCMModelName.txt
  TABLE_RCM_NAME=CORDEX_RCMs_ToU.txt

# Specification of requirements.
# An automatically updated version by default.
  TABLE_VAR_REQ=CORDEX_variables_requirement_table.csv

# By default, the record of time is tested on the basis of regular
# steps and for consistency across files and experiments. However
# the entire set of CMIP5 experiments includes also irregularities
# and distinct periods. Time information was extracted from the
# CMIP5 standard table and is given in QA-DKRZ/tables/projects/CMIP5/time_table.csv
# which provides a default. Please have a copy of your own with a
# different name in case of individual settings; the file
# 'time_table.csv' is part of svn updates. Please have a very close
# look at the file, because this table is certainly not free of errors.
# If you detect any, please give a short note.
# The default falls back to 'regular' when no time table file is
# found in QA-DKRZ/tables/projects/CMIP5.
# TABLE_TIME_SCHEDULE=...  [time_table.csv]


# Section: Command-line used Options
# ----------------------------------

# Remove former results corresponding to the paths and variables
# selected before re-doing the QA. However, corresponding logfiles
# are not affected. Assignments:
# 1) CLEAR=only, do a clearing, but no QA will take place.
# 2) CLEAR=lock for files qc_lock_<filename>.txt .
# 3) CLEAR=note for qc_note_<filename>.txt.
# 4) CLEAR='resume', then qc_lock_<filename>.txt files will
#    be removed before processing any further.
# 5) CLEAR=follow_links. When targets are pointed to
#    by symbolic links, then both target and link will be removed.
#    Note: symbolic links are removed by default, but not the targets.
# 6) CLEAR=mark, when a file 'clear.mark' or a file
#    <filename_root>.clear is present.
# Multiple options by a comma-separated-list; the order doesn't matter.
# Note: CLEAR=lock,note or mark enables option FORCE_STND_TABLE for the respective variable.
# CLEAR=only
# CLEAR=lock,note

# Check and show availability of tools and exit.
# Note that this is done at the beginning of each session.
# CHECK_TOOLS

# Log execution of qcConfigurator processing to STDERR.
# DEBUG_C

# Log execution of qcExecutor processing to files corresponding to
# variable names (in the directory where qcManager was started).
# DEBUG_E

# Log execution of the qcManager to STDERR.
# DEBUG_M

# Enable debugging of external scripts invoked in qcManager; default: every.
# If the name of a script is specified, then this applied only to this one.
# DEBUG_X[=script]

# Flow trace analysis of the qcManager main-loop.
# Note that specific commenting-in is required for qcManager and qcExecutor.
# e.g.: sed -i 's/#trace/trace/' qcManager
# FLOW_TRACE

# Move/merge QA results, when the corresponding data have been moved.
# a) slash-less value: append this to all paths of all experiments affected
#    by the configuration selection.
# MOVE_LOG=v12345

# Process the next num variable-frequency occurrences.
# NEXT[=num]

# Only progress by the given number of records at each check.
# Purpose: probably debugging
# NEXT_RECORDS=50

# Results are not summarised.
# NO_SUMMARY

# Force creation of annotation files in QA_RESULTS/check_logs.
# Usually for testing, because results are summarised at the end of each run.
# If a comma-separated list of experiment names is assigned to the option,
# then only these are summarised. Else, the experiment names resulting from
# the current configuration in use.
# ONLY_SUMMARY  # [ disabled]

# Show the qC_main.x call; no execution. Purpose: debugging of C++ programs
# SHOW_CALL

# Show files whose qa results would be affected by a clearance.
# SHOW_CLEAR

# Show configuration and exit.
# SHOW_CONF

# Show paths affected by the current SELECTion and the corresponding log-file name.
# SHOW_EXP

# Show the names of log-files created or modified by the current SELECTion.
# SHOW_LOG_FNAME

# Show the next num variable-frequency occurrences scheduled for processing.
# SHOW_NEXT[=num]


# Section: Processing
# -------------------

# Automatic update of the QA package. If enabled, automatic update
# is kept until it is disabled. Alternatively and easier is the
# command-line option '--auto-up' for the scripts
# 'qcManager' and 'install'. Keword '--auto-up' enables while
# '--auto-up=disable' cancels any previous setting.
# AUTO_UPDATE[=disable] # [disabled]

# This option affects only the test of periods i the filename vs. the
# span of time values given in the file.
# DATA_IN_PRODUCTION

# Send a summary of notifications and checked periods of variables
# to this (list of) email recipients after each finished QA session.
# Default: no email is sent.
# Note: this is different from EMAIL_TO.
# EMAIL_SUMMARY=name@site.dom [disabled]

# In case of error: send e-mails to this comma-separated-list.
# EMAIL_TO=name@site.dom #[disabled]

# If the QA stores results in a files system with occasional
# 'no space left on device', then you could enable a disk space
# inquiry before write operations are started for each variable.
# ENABLE_DISK_SPACE_INQ  # [disabled]

# Set group name. This is necessary, if the QA is also operated
# by users who are not in the default group of the user who initially
# checked out from the svn repository. This will automatically also
# set the SGID-bit, i.e. grant permissions to all group members.
# The current setting of user permissions of each file is duplicated
# to the respective group permissions.
# Note: this can only be done by the owner of the QA directory.
# GROUP_NAME=  # [none]

# In order to enable trapping signals, long sleeping period are
# subdivided into smaller intervals of consecutive sleep commands.
# HARD_SLEEP_PERIOD=10  # [10]s

# If there are (temporary) files failing the right filename syntax
# required for temporal sub-files AND it is obvious that the syntax
# rules are always followed, then ignore these in the syncFile process.
# Exception: files below the 'fx' directory in the data directory tree.
# IGNORE_TEMP_FILES

# Available disk space is checked before a file is processed.
# If you know that there will be always space enough, then enable
# this option. NOTE: If the compiler complains about
# structure 'statvfs64', please try the macro -DSTATVFS parameter
# for compilation of the program 'diskUsage.c'.
# INQ_DISK_SPACE  # [disabled]

# Notifications of level 4 as well as segmentation fault will terminate
# a running session immediately. This behaviour may be cancelled.
# Note that Ctrl C, i.e. TERM, terminates nevertheless always.
  NEVER_BREAK_SESSION

# Apply linux 'nice' command to executables on
# guest machines (see 'man nice').
# NICE=15  # [0]

# Suspend processes between 8 - 19 o'clock on guest hosts and
# allow only a single process on the qcManager host.
# NIGHT_SHIFT # [disabled]

# List of number(s) of simultaneous execution processes per host.
# Note: there is always a single qcManager process.
# Special: If a single number is specified, this is assigned to
# all hosts in the QA_EXEC_HOSTS list.
# Fine tuning: each positional number in the list corresponds to a
# position in the QA_EXEC_HOSTS list. If less positions are given
# than in QA_EXEC_HOSTS, then the last position is assigned to
# the omitted positions.
  NUM_EXEC_THREADS=2 #1,0  # [1]

# More configuration options in another file. These options have a lower priority than those
# given in the current file. Note that there is no nesting depths
# of additional configuration files.
# QA_CONF=another-configuration-file

# The list of machines executing jobs for this session. Usually disabled
# causing the current computer to be the only one.
# QA_EXEC_HOSTS=comma-separated-list-of-hosts  # [HOSTNAME of qcManager]

# If a process cannot be run, because  the data base is not available
# or a server is absent for instance, then retry after a sleep for the specified number.
# REATTEMPT_LIMIT=5  # [5]

# The traditional processing of the summary was done by bash scripts; at present,
# however, by QA_DKRZ/scripts/python/qa-dkrz/qa_summary.py. If new results
# are appended to QA_RESULTS, the program recognises traditional or new
# and switches respectively. Enabling TRADITIONAL_SUMMARY and PY_SUMMARY
# will force correspondingly.
# TRADITIONAL_SUMMARY  # [disabled]
# PY_SUMMARY  #[yes]

# Sleep period if waiting for a server or anything else. Note:
# particular ordinary processes have a sleep time of their own
# to distinguish them in the output of 'ps'.
# Start the analysis on a guest machine, if the average load from
# the uptime command is < 1.5 for 5 minutes and < 1 for 15 min.
# This option is not available at present.
# WORK_AT_LOW_LOAD  # [disabled]

# Kill a stalled check-job (not the session) after specified seconds.
# Note that such a kill would usually initiate a mail message and would
# generate a lock-file.
# ZOMBIE_LIMIT=3600  # [3600]

# Section: Miscellaneous
# ---------------------

# By default, calculation of cell averages over a geographical grid is
# performed by weighting with the corresponding grid-cell area. In
# particular for ocean basin cross-section, this is very time consuming
# due to the irregular shape of the basins taken into account by
# varying filling values (also for regular grids with varying effective
# areas, e.g. T of snow). If the precise value of the average is less
# important, this option reduces calculation time.
  ARITHMETIC_MEAN  # [disabled]

# The QA checks are organised in modules, which can be selected
# independently by a comma-separated, case-independent list [all by default].
# Modules: {DRS, DRS_F, DRS_P, CF, CV, CNSTY}=META, TIME, and DATA.
# Acronyms: Data Reference System (DRS) which could be specialised by DRS_F for a
# filename check and DRS_P for the path, Climate Forcing Convention (CF),
# Controlled Vocabulary (CV), and consistency check (CNSTY).
# It is also possible to disable a mode by putting a minus sign in front of a key .
# CHECK_MODE=meta  #[all]

# Calculate checksum of every file that passes successfully the QA.
# md5sum is applied by default, but sha1sum as well as any other program
# giving a checksum as first item in the output will do; abbreviation sha1
# will also work as assignment to CHECKSUM. Any other program must be available in
# the user's search path for commands or specified with absolute path.
# Checksum, filename, path components, and modification time of the file are
# written to a checksum-table; creation_date and tracking_id are added when available.
# If any of the CS_* options below is enabled, then CHECKSUM or CHECKSUM=t may be discarded.
# CHECKSUM  # [disabled]

# Path to the checksum-table; by default in QA_RESULTS/cs_table.
# CS_DIR=cs_table

# Checksums are unique, and creations_date and tracking_id are supposed just as well.
# Thus, when a file changed, but creation_date or tracking_id did not, will be annotated.
# This option ignores earlier values.
  CS_NO_HISTORY

# Checksums are stored in container-files by default. This option lets store checksum etc.
# for each file separately in the QA_RESULTS/data/DRS-PATH with a corresponding filename. If
# CS_DIR is also provided, then the checksum-files are stored therein.
# CS_STAND_ALONE

# If there is no definition of variables's attributes valid_range attribute,
# valid_max, valid_min, respectively, then these defaults are used.
  DEFAULT_VALID_MAX=1E+17
  DEFAULT_VALID_MIN=-1E-15

# Disable testing for Inf or Nan.
# Note that this may also be disabled by setting D[iscard] in the check-list file
# for flag R6400.
# DISABLE_INF_NAN

# Exclude specific attributes from any check.
# EXCLUDE_ATTRIBUTE=history,comment,associated_files

# Exclude specific variables globally from a check.
# EXCLUDE_VARIABLE=average_T1,average_T2,average_DT

# Issue a note for every variable-frequency case specified in
# CORDEX_variables_requirement, which is not available in a data set,
# to files annotation_outline.txt and annotation_summary.txt
# within the QA RESULTS/check_logs/_Notes directory.
# FIND_NOT_DELIVERED [enabled}

# By default, the reference date of variable 'time' written to the
# project table is checked for each file, even across experiments.
# There is nothing wrong with different reference dates in principle
# thus throwing warnings can be disabled. Please, notice the difference
# to the option IGNORE_REFERENCE_DATE, which disables any check for
# differing reference dates, even across files.
  IGNORE_REF_DATE_ACROSS_EXP  # [ disabled]

# This option disables any check for
# differing reference dates, even across files
# (includes option IGNORE_REF_DATE_ACROSS_EXP).
  IGNORE_REFERENCE_DATE  # [disabled]

# Log execution time of the QA.cpp exectuable
# LOG_CPU_TIME  # [disabled]

# This disables any check related to a constant time step within a
# file, across files, or across experiments. In fact, the only
# property of time to be checked is that time marches on and that
# time bounds don't overlap.
# NON_REGULAR_TIME_STEP  # [disabled]

# User-defined handling of exceptions, i.e. how to proceed with
# notification of QA check results. Explanations in file
# QA-DKRZ/projects/CORDEX/CORDEX_check-list.conf.
# Keys: Flag, L[1-4], D, PT, ST, variable names. Please note that
# any misspelling is considered the name of a variable and would
# throw a flag in most cases.
# NOTE={o2,PT,58_2,L1}

# Apply a unique annotation handling to all entries of the check-list.
# NOTE_ALWAYS=L1,EM

# Limit maximum annotation level to all entries of the check-list.
# NOTE_LEVEL_LIMIT=L1

# Min/max of data are tested whether they could be out of scope
# of reasonable value ranges. Peaks/breaks are considered such that
# a constant number of data points satisfys M-times of n in a row the condition
#    abs( value - average) > ( stdDev + n * stdDev / 4 )
# with increment n= 1,2,3,... Average, standard deviation, and range
# for the statistics of global minimum and maximum, respectively.
# But, if more than P % of all the data points were found to be outside,
# then these are considered regular. In that case, outliers are
# still detected when exceeding orders of magnitude L. Alternatively
# a fixed number N, which must not be exceeded, may be provided. Both
# N and P are accepted and taken into account simultaneously.
# Default setting: M=5/P=1/L=5, disabled N);
# M,P, and/or L may be assigned to the OUTLIER_TEST option.
# Keyword 'ignore' suppresses email notification as usual.
# Comma-sep-list of variables may be specified. Groups of parameters may be
# indicated by braces {}, which may be nested.
# Multiple lines are accepted, even with unclosed braces. All lines
# will be concatenated in the order of appearance.
# Matching pairs of '{' and '}' are checked.
# Compatibility: ignore{var_1,...,var_n} still accepted (same for discard).
# OUTLIER_TEST  # [def: enabled with M5,P0.01,L5 ]
  OUTLIER_TEST=N10,P0.01,M3,L3

# The name of the project. Supported at present: CMIP5, CORDEX.
# If project of different name follows (almost) a supported one,
# then enabling option 'NOTE={2_8,D}' discards mis-match annotations.
# Disabling suppresses related meta-data checks.
  PROJECT=CORDEX # [CMIP5]

# Switch on a Quality Control run (on by default). It is possible to
# switch it off, for instance to calculate only frequency distributions.
# (not a good example, because both can be run simultaneously).
# QA=off  # [on]

# Find occurrence of multiple identical data records.
# A checksum of all data is calculated for each record and compared
# to each other. Special assignments (without ''):
# 1) no_0-D:      ignore for 0-D variables.
# 2) only_groups: omit isolated replicated records, report only groups.
#                   The size of the group may be specified by appending
#                   '=num'; by default num=2.
# 3) clear_bitsN: clear the N least significant bits of data values
#                   with default of N=2.
# REPLICATED_RECORD=pr,tas
# REPLICATED_RECORD=no_0-D,only_groups_12,clear_bits
# REPLICATED_RECORD=no_0-D,only_groups=12

# Stop the QA at a specific time value.
# Use ISO-8601 format (yyyy-mm-ddThh:mm:ss). Truncations are accepted
# (1955-06 would process the records before 1955-06-01 00:00:00).
# If the parameter gets a leading 't' or 'T' character, then
# a time value (a pure number) is accepted that must correspond to the
# reference date in the data file.
# TIME_LIMIT=1955-08


# Section: Frequency Distribution
# -------------------------------
# Please, note thath this was not tested for a rather long time.

# Calculate the frequency distribution of the entire domain
# Note: generally, you will need post-processing with 'hist.x' to
# assemble the final frequency distribution, except you select specific
# debug options.
# Rules for selecting class widths and starting point (e.g. centering):
# Firstly, a fd-build-file of the same variable at destination has precedence.
# Second choice is a fd-build-file or fd-prop-file (the header of
# a build-file) in a specified path.
# Thirdly, explicit statement of properties.
# Default: automatic determination of fd properties.
# FREQ_DIST  # [disabled]

# Partitioning the total time span of an experiment into
# smaller ranges. FreqDists will be calculated separately for each
# window. Note: the residual time interval may be shorter.
# Note: if no unit-designator is appended (y, mo, d, h, mi, s)
# the unit from the time variable is used.
# FD_TIME_PART=5y # 30y [ default: the entire experiment]

# By default, frequency distributions adjust automatically the bin (class) width
# while processing the first time window (or the total time span),
# which is then used in subsequent time windows for the entire experiment.
# Specifying this path enables to retain the properties of the same
# variable (if found) of another experiment, e.g. control run.
# Such a file must have extension .build or .prop
# FD_PROPERTY_PATH=  # [no path]

# Explicit properties (two quantities separated by '/'):
#    class-width (number) / init value (number),
# e.g. 1/0 will centre a bin (class) of the freq.dist around zero.
# Note: The alignment of the fd-bin borders is done automatically.
#       It is not possible, do have a class boundary exactly
#       at say 273.15 with class-width of 1.
# This option is only reasonable for selecting a particular common type
# of variable, e.g. temperature, in a separate run. See priority rules.
# FD_EXPLICIT_PROPS=1.E-05/3.5E-05

# Output of the Freq Dist as ready-to-use. By default, the output
# will be in a format to be re-read later in order to resume a
# previous session (the output file gets extension '.build').
# File extension in plain mode: '.hist'. Attention: this starts
# and completes a FD calculation. Makes little sense for
# multi-sessions. Purpose: debugging
# FD_PLAIN  # [disabled]

# Like FREQ_DIST_PLAIN, but output of the Freq Dist shows complete
# shapes of bars. Purpose: debugging
# FD_BARS  # [disabled]

# Section: Syntax Rules of the Configuration File
# -----------------------------------------------
# KEY WORDS controlling the application in the quality-control-run.
# SYNTAX: KEY-WORD[[+][=]value[,value[,value]]
#

# -) KEY-WORD
#    Consisting only of alphanumeric characters and underscore;
#    only upper-case with alphabetic begin. KEY_WORD and following
#    '=' or '+=' may be separated by spaces.
# -) KEY-WORD
#    Set alone enables a corresponding feature. Equivalent to KEY-WORD=t .
# -) KEY-WORD=value
#    Make a simple assignment; any previous assignment is overwritten;
#    also by KEY-WORD alone.
# -) KEY-WORD=value1,value2, ...
#    Assign a comma-separated-expression. This will be expanded to an array.
# -) KEY-WORD+=value
# -) Add another item to a comma-separated expression.
# -) Configuration options may be supplied on the command line with the prefix -E_.
# -) Options SELECT and LOCK are special.
#    Additionally, -S param and -L param may be given alternatively
#    on the command-line.
# -) Invalid/misspelled KEY_WORD is discarded without notice;
#    this may cause 'amazement'.
#    The effect of invalid assignments is trapped sometimes.

# Note for SELECT, LOCK, -S and -L usage:
# General setting:  {space, = ,+=}[path[=]][variable]

# Examples:
# leading character(s) supplied by the calling instances
# -) '=': define or overwrite
# -) '+=': prolong array of paths/variables
# -) no leading '=' character: equivalent to +=

# -) path=var
#    specifies a single path where to look for a single variable
# -) p1,p2=var
#    two paths to look for a variable
# -) p1,p2=
#    two paths with every variable, equivalent to p1,p2=.*
# -) p1=v1,v2
#    one path with two variables
# -) str1[,str2,str3]
#    a) no '/' --> variables
#    b) '/' anywhwere  --> only paths; no trapping of errors
# -) relative path without a '/' must have '=' appended
# -) p1=,p2=v1...  --> ERROR