Configuration Metadata

Configuration metadata uses the standard Rose Configuration Format. It is represented in a directory with the following:

  • rose-meta.conf, the main metadata configuration file.

  • opt/ directory (see Optional Configuration).

  • Other files, e.g.:

    • lib/python/widget/ would be the location of a custom widget.

    • lib/python/macros/ would be the location of a custom macro.

    • etc/ would contain any resources for the logics in lib/python/, such as an icon for the custom widget.

Rose utilities will search for metadata using the following in order of precedence:

  1. Configuration metadata embedded in the meta/ directory of a suite or an application.

  2. The --meta-path=PATH option of relevant commands.

  3. The value of the ROSE_META_PATH environment variable.

  4. The rose.conf|meta-path setting (see Site And User Configuration).


See Appendix: Metadata Location for more details.

The configuration metadata that controls default behaviour will be located in etc/rose-meta/ within the metomi.rose Python installation.

Configuration Metadata File

The rose-meta.conf contains a serialised data structure that is an unordered map (sections=) of unordered maps (keys=values).

The section name in a configuration metadata file should be a unique ID to the relevant configuration setting. The syntax of the ID is SECTION-NAME=OPTION-NAME or just SECTION-NAME. For example, env=MY_ENV_NAME is the ID of an environment variable called MY_ENV_NAME in an application configuration file; namelist:something_nl=variable_name1 is the ID of a namelist variable called variable_name1 in a namelist group called something_nl in an application configuration file. If the configuration metadata applies to a section in a configuration file, the ID is just the section name.

Where multiple instances of a section are used in a configuration file, ending in brackets with numbers, the metadata ID should just be based on the original section name (for example namelist:extract_control(2) should be found in the metadata under namelist:extract_control).

Finally, the configuration metadata may group settings in namespaces, which may in turn have common configuration metadata settings. The ID for a namespace set in the metadata is ns=NAME, e.g. ns=env/MyFavouriteEnvironmentVars.

Metadata Inheritance (Import)

A root level import=MY_COMMAND1/VERSION1 MY_COMMAND2/VERSION2 ... setting in the rose-meta.conf file will tell Rose metadata utilities to locate the meta keys MY_COMMAND1/VERSION1, MY_COMMAND2/VERSION2 (and so on) and inherit their configuration and files if found.

For example, you might have a rose-meta directory that contains the following files and directories:


and write an app referencing this rose-meta directory that looks like this:



This will reference the metadata at rose-meta/cheese_sandwich/vn2.0.

Now, we can write the rose-meta.conf file using an import:



which will inherit metadata from metadata from rose-meta/cheese/vn1.0/rose-meta.conf.

Metadata Options

The metadata options for a configuration fall into four categories: sorting, values, behaviour and help as outlined below.

Rose Configuration rose-meta.conf
Config [SETTING]

A section containing metadata items relating to a particular setting.

SETTING should be the full name of a configuration containing the name of the section and the name of the setting separated by an equals = sign e.g:

  • [env=FOO] would refer to the environment variable FOO

  • [namelist:foo=BAR] would refer to BAR from the namelist foo.

Metadata for Sorting

These configuration metadata are used for grouping and sorting the IDs of the configurations.

Config ns

A forward slash / delimited hierarchical namespace for the container of the setting, which overrides the default. The default namespace for the setting is derived from the first part of the ID - by splitting up the section name by colons : or forward slashes /. For example, a configuration with an ID namelist:var_minimise=niter_set would have the namespace namelist/var_minimise. If a namespace is defined for a section, it will become the default for all the settings in that section.

The namespace is used by rose config-edit to group settings, so that they can be placed in different pages. A namespace for a section will become the default for all the settings in that section.


You should not assign namespaces to variables in duplicate sections.

Config sort-key

A character string that can be used as a sort key for ordering an option within its namespace.

It can also be used to order sections and namespaces.

The sort-key is used by rose config-edit to group settings on a page. Items with a sort-key will be sorted to the top of a name-space. Items without a sort-key will be sorted after, in ascending order of their IDs.

The sorting procedure in pseudo code is a normal ASCII-like sorting of a list of setting_sort_key + "~" + setting_id strings. If there is no setting_sort_key, null string will be used.

For example, the following metadata:






would produce a sorting order of env=prune, env=melon, env=cherry, env=apple, env=banana.

Metadata for Values

These configuration metadata are used to define the valid values of a setting. A Rose utility such as rose config-edit can use these metadata to display the correct widget for a setting and to check its input. However, if the value of a setting contains a string that looks like an environment variable, these metadata will normally be ignored.

Config type


The type/class of the setting. The type names are based on the intrinsic Fortran types, such as integer and real. Currently supported types are:


example option: PRODUCE_THINGS=true

description: either true or false

usage: environment variables, javascript/JSON inputs


example option: sea_colour='blue'

description: Fortran character type - a single quoted string, single quotes escaped in pairs

usage: Fortran character types


example option: num_lucky=5

description: generic integer type

usage: any integer-type input


example option: l_spock=.true.

description: Fortran logical type - either .true. or .false.

usage: Fortran logical types


example option: ENABLE_THINGS=True

description: Python boolean type - either True or False

usage: Python boolean types


description: used to signify a Python-compatible formatted list such as ["Foo", 50, False].


This encapsulates length, so do not use a separate length declaration for this setting.

usage: use for inputs that expect a string that looks like a Python list - e.g. Jinja2 list input variables.


example option: js_measure_cloud_mode="laser"

description: a double quoted string, double quotes escaped with backslash

usage: Inputs that require double quotes and allow backslash escaping e.g. javascript/JSON inputs.


example option: n_avogadro=6.02e23

description: Fortran real number type, generic floating point numbers

usage: Fortran real types, generic floating point numbers.


Scientific notation must use the “e” or “E” format.

comment: Internally implemented within Rose using Python’s floating point specification.


description: placeholder used in derived type specifications where none of the above types apply

usage: only in derived types


description: for strings containing newline characters.

usage: plain text strings


description: used to signify a space separated list such as "Foo" 50 False.

usage: use for inputs that expect a string that contains a number of space separated items - e.g. in fcm_make app configs.


Not all inputs need to have type defined. In some cases using values or pattern is better.

A derived type may be defined by a comma , separated list of intrinsic types, e.g. integer, character, real, integer. The default is a raw string.

Config length

Define the length of an array. If not present, the setting is assumed to be a scalar. A positive integer defines a fixed length array. A colon : defines a dynamic length array.


You do not need to use length if you already have type=python_list for a setting.

Config element-titles

Define a list of comma separated “titles” to appear above array entries. If not present then no titles are displayed.


Where the number of element-titles is greater than the length of the array, it will only display titles up to the length of the array. Additionally, where the associated array is longer than the number of element-titles, blank headings will be placed above them.

Config values

Define a comma , separated list of permitted values of a setting (or an element in the setting if it is an array). This metadata overrides the type, range and pattern metadata.

For example, rose config-edit may use this list to determine the widget to display the setting. It may display the choices using a set of radio buttons if the list of values is small, or a drop down combo box if the list of values is large. If the list only contains one value, rose config-edit will expect the setting to always have this value, and may display it as a special setting.

Config value-titles

Define a comma , separated list of titles to associate with each of the elements of values which will be displayed instead of the value. This list should contain the same number of elements as the values entry.

For example, given the following metadata:

values=0, 1, 2
value-titles=low, medium, high

rose config-edit will display low for option value 0, medium for 1 and high for 2.

Config value-hints

Define a comma , separated list of suggested values for a variable as value “hints”, but still allows the user to provide their own override. This is like an auto-complete widget.

For example, given the following metadata:


rose config-edit will display possible option values when the user starts typing if they match a suggested value.

Config range

Specify a range of values. It can either be a simple comma , separated list of allowed values, or a logical expression in the Rose metadata mini-language. This metadata is only valid if type is either integer or real.

A simple list may contain a mixture of allowed numbers and number ranges such as 1, 2, 4:8, 10: (which means the value can be 1, 2, between 4 and 8 inclusive, or any values greater than or equal to 10.)

A logical expression uses the Rose metadata mini-language, using the variable this to denote the value of the current setting, e.g. this <-1 and this >1.


Inter-variable comparisons are not permitted (but see the fail-if property below for a way to implement this).

Config pattern

Specify a regular expression (Python Regular Expressions) to compare against the whole value of the setting.

For example, if we write the following metadata:


then we expect all valid values for country_of_origin to start with a double quote (^"), begin with an uppercase letter ([A-Z]), contain some other characters or spaces (.+), and end with a quote ("$).

If you have an array variable (for example, TARTAN_PAINT_COLOURS='blue','red','blue') and you want to specify a pattern that each element of the array must match, you can construct a regular expression that repeats and includes commas. For example, if each element in our TARTAN_PAINT_COLOURS array must obey the regular expression 'red'|'blue', then we can write:

Config fail-if

Specify a logical expression using the Rose mini-language to validate the value of the current setting with respect to other settings. If the logical expression evaluates to true, the system will consider the setting in error.

See the associated setting warn-if for raising warnings.

The logical expression uses a Python-like syntax (documented in the appendix). It can reference the value of the current setting with the this variable and the value of other settings with their IDs. E.g.:

fail-if=this < namelist:test=control_lt_var;

means that an error will be flagged if the numeric value of my_test_var is less than the numeric value of control_lt_var.

fail-if=this != 1 + namelist:test=ctrl_var_1 * (namelist:test=ctrl_var_2 - this);

shows a more complex operation, again with numeric values.

To check array elements, the ID should be expressed with a bracketed index, as in the configuration:

fail-if=this(2) != "'0A'" and this(4) == "'0A'";


With array elements indexing starts from 1.

Array elements can also be checked using any and all. E.g.:

fail-if=any(namelist:test=ctrl_array < this);
fail-if=all(this == 0);

Similarly, the number of array elements can be checked using len. E.g.:

fail-if=len(namelist:test=ctrl_array) < this;
fail-if=len(this) == 0;

Expressions can be chained together and brackets used:

fail-if=this(4) == "'0A'" and (namelist:test=ctrl_var_1 != "'N'" or
namelist:test=ctrl_var_2 != "'Y'" or all(namelist:test=ctrl_arr_3 == 'N'));

Multiple failure conditions can be added, by using a semicolon as the separator - the semicolon is optional for a single statement or the last in a block, but is recommended. Multiple failure conditions are essentially similar to chaining them together with or, but the system can process each expression one by one to target the error message.

fail-if=this > 0; this % 2 == 1; this * 3 > 100;

You can add a message to the error or warning message to make it clearer by adding a hash followed by the comment at the end of a configuration metadata line:

# Need common divisor for ctrl_array
fail-if=any(namelist:test=ctrl_array % this == 0);

When using multiple failure conditions, different messages can be added if they are placed on individual lines:

fail-if=this > 0; # Needs to be less than or equal to 0
        this % 2 == 1; # Needs to be odd
        this * 3 > 100; # Needs to be more than 100/3.


When dividing a real-numbered setting by something, make sure that the expression does not actually divide an integer by an integer - for example, this / 2 will evaluate as 0 if this has a value of 1, but 0.5 if it has a value of 1.0. This is a result of Python’s implicit typing.

You can get around this by making sure either the numerator or denominator is a real number - e.g. by rewriting it as this / 2.0 or 1.0 * this / 2.

Config warn-if

Specify a logical expression using the Rose mini-language to validate the value of the current setting with respect to other settings. If the logical expression evaluates to true, the system will issue a warning. It is a slightly different usage of the fail-if functionality which can do things like warn of deprecated content, e.g.:


would always evaluate True and give a warning if the setting is present.

See the associated setting fail-if for examples of logical expressions that may be added.

Metadata for Behaviour

These metadata are used to define how the setting should behave in different states.

Config compulsory

A true value indicates that the setting should be compulsory. If this flag is not set, the setting is optional.

Compulsory sections should be physically present in the configuration at all times. Compulsory options should be physically present in the configuration if their parent section is physically present.

Optional settings can be removed as the user wishes. Compulsory settings may however be triggered ignored (see trigger). For example, the rose config-edit may issue a warning if a compulsory setting is not defined. It may also hide a compulsory variable that only has a single permitted value.

Config trigger

A setting has the following states:

  • normal

  • user ignored (stored in the configuration file with a ! flag, ignored at runtime)

  • logically ignored (stored in the configuration file with a !! flag, ignored at runtime)

If a setting is user ignored, the trigger will do nothing. Otherwise:

  • If the logic for a setting in the trigger is fulfilled, it will cause the setting to be enabled.

  • If it is not, it will cause the setting to be logically ignored.

The trigger expression is a list of settings triggered by (different values of) this setting. If the values are not given, the setting will be triggered only if the current setting is enabled.

The syntax contains ID-value pairs, where the values part is optional. Each pair must be separated by a semi-colon. Within each pair, any values must be separated from the ID by a colon (:) and a space. Values must be formatted in the same way as the setting values defined above (i.e. comma separated).

The trigger syntax looks like:

        namelist:value_nl=X: 10;
        env=Y: 20, 30, 40;
        namelist:value_nl=Z: 20;

In this example:

  • When namelist:trig_nl=trigger_variable is ignored, all the variables in the trigger expression will be ignored, irrespective of its value.

  • When namelist:trig_nl=trigger_variable is enabled, namelist:dep_nl=A and namelist:dep_nl=B will both be enabled, and the other variables may be enabled according to its value:

    • When the value of the setting is not 10, 20, 30, or 40, namelist:value_nl=X, env=Y and namelist:value_nl=Z will be ignored.

    • When the value of the setting is 10, namelist:value_nl=X will be enabled, but env=Y and namelist:value_nl=Z will be ignored.

    • When the value of the setting is 20, env=Y and namelist:value_nl=Z will be enabled, but namelist:value_nl=X will be ignored.

    • When the value of the setting is 30, env=Y will be enabled, but namelist:value_nl=X and namelist:value_nl=Z will be ignored.

    • If the value of the setting contains an environment variable-like string, e.g. ${TEN_MULTIPLE}, all three will be enabled.

Settings mentioned in trigger expressions will have their default state set to ignored unless explicitly triggered. rose config-edit will issue warnings if variables or sections are in the incorrect state when it loads a configuration. Triggering thereafter will work as normal.

Where multiple triggers act on a setting, the setting is triggered only if all triggers are active (i.e. an AND relationship). For example, for the two triggers here:

trigger=env=IS_ICE: true;

trigger=env=IS_ICE: true;

the setting env=IS_ICE is only enabled if both env=IS_WATER and env=IS_COLD are true and enabled. Otherwise, it is ignored.

The trigger syntax also supports a logical expression using the Rose metadata mini-language, in the same way as the range or fail-if metadata. As with range, inter-variable comparisons are disallowed.

trigger=env=CUSTOM_SNOWFLAKE_GEOMETRY: this != 6;
        env=SILLY_SNOWFLAKE_GEOMETRY: this < 2;

In this example, the setting env=CUSTOM_SNOWFLAKE_GEOMETRY is enabled if env=SNOWFLAKE_SIDES is enabled and not 6. env=SILLY_SNOWFLAKE_GEOMETRY is only enabled when env=SNOWFLAKE_SIDES is enabled and less than 2. The logical expression syntax can be used with non-numeric variables in the same way as the fail-if metadata.

It is possible to use metadata triggers to trigger file creation, switching on/off as required.

For example, with the following rose-app.conf and rose-meta.conf files, file creation is triggered when setting trigger=file:foo to .true..

rose-app.conf file:




rose-meta.conf file:


trigger=file:foo: .true.
Config duplicate

Allow duplicated copies of the setting. This is used for sections where there may be more than one with the same metadata - for example multiple namelist groups of the same name. If this setting is true for a given name, the application configuration will accept multiple namelist groups of this name. rose config-edit may then provide the option to clone or copy a namelist to generate an additional namelist. Otherwise, rose config-edit may issue warning for configuration sections that are found with multiple copies or an index.

Config macro

Associate a setting with a comma-delimited set of custom macros (but not upgrade macros).

E.g. for a macro class called FibonacciChecker in the metadata under lib/python/macros/, we may have:


This may be used in rose config-edit to visually associate the setting with these macros. If a macro class has both a transform and a validate method, you can specify which you need by appending the method to the name e.g.:

Config widget[gui-application]

Indicate that the gui-application (e.g. rose config-edit) should use a special widget to display this setting.

E.g. If we want to use a slider instead of an entry box for a floating point real number.

The widget may take space-delimited arguments which would be specified after the widget name. E.g. to set up a hypothetical table with named columns X, Y, VCT, and Level, we may do:

widget[rose-config-edit]=table.TableWidget X Y VCT Level

You may override to a Rose built-in widget by specifying a full rose class path in Python - for example, to always show radiobuttons for an option with values set:


Another useful Rose built-in widget to use is the array element aligning page widget, metomi.rose.config_editor.pagewidget.table.PageArrayTable. You can set this for a section or namespace to make sure each n-th variable value element lines up horizontally. For example:

main='beef','spaghetti','coq au vin','lamb'

could use the following metadata:


to align the elements on the page like this:

customers        Athos      Porthos      Aramis      d'Artagnan
entrees          soup        pate         soup       asparagus
main             beef      spaghetti   coq au vin       lamb
petits_fours    .false.     .true.       .false.       .true.
Config copy-mode

For use with settings in the file.

Setting copy-mode in the metadata allows for the field to be either never copied or copied with any value associated to be clear.

For example: in a file:

[ensemble members]

Setting the ensemble members field to include copy-mode=never means that the ensemble members field would never be copied.

[additional info]

Setting the additional info field to include copy-mode=never means that the additional info field would be copied, but any value associated with it would be cleared.

Metadata for Help

These metadata provide help for a configuration.

Config url

A web URL containing help for the setting. For example:


For example, the rose config-edit will trigger a web browser to display this when a variable name is clicked. A partial URL can be used for variables if the variable’s section or namespace has a relevant parent url property to use as a prefix. For example:


Config help

(Long) help for the setting. For example, rose config-edit will use this in a pop-up dialog for a variable. Embedding variable IDs in the help string will allow links to the variables to be created within the pop-up dialog box, e.g.

help=Used in conjunction with namelist:Var_DiagnosticsNL=n_linear_adj_test to do something linear.

Web URLs beginning with http:// will also be presented as links in the rose config-edit.

Config description

(Medium) description for the setting. For example, rose config-edit will use this as part of the hover over text.

rose config-edit will also use descriptions set for sections or namespaces as page header text (appears at the top of a panel or page), with clickable ID and URL links as in help. Descriptions set for variables may be automatically shown underneath the variable name in rose config-edit, depending on view options.

Config title

(Short) title for the setting. For example, rose config-edit can use this specification as the label of a setting, instead of the variable name.

Appendix: Metadata Location

Centralised Rose metadata is referred to with either the rose-suite.conf|meta or|project settings in a suite configuration. It needs to live in a system-global-readable location.

Rose utilities will do a path search for metadata using the following in order of precedence:

Each of the above settings can be a colon-separated list of paths.

Underneath each directory in the search path should be a hierarchy like the following:

${APP}/ # i.e. the upgrade macros


A is likely to have no versions.


In some cases, a number of different executables may share the same application configuration metadata in which case APP is given a name which covers all the uses.


The Rose team recommend placing metadata in a rose-meta directory at the top of a project’s source tree. Central metadata, if any, at the meta-path location in the site configuration, should be a collection of regularly-updated subdirectories from all of the relevant projects’ rose-meta directories.

For example, a system CHOCOLATE may have a flat metadata structure within the repository:


and the system CAFFEINE may have a hybrid structure, with both flat and hierarchical components:


and a site configuration with:


We would expect the following directories in /path/to/rose-meta:


with caffeine-coffee containing subdirectories cappuccino and latte, and caffeine-tea containing yorkshire and lapsang.

Upgrade and Versions


The HEAD (i.e. development) version

The configuration metadata most relevant to the latest revision of the source tree.

A named version

The configuration metadata most relevant to a release, or a particular revision, of the software. This will normally be a copy of the HEAD version at a given revision, although it may be updated with some backward compatible fixes.

Each change in the HEAD version that requires an upgrade procedure should introduce an upgrade macro. Each upgrade macro will provide the following information:

  • A tag of the configuration which can be applied by this macro (i.e. the previous tag).

  • A tag of the configuration after the transformation.

This allows our system to build up a chain if multiple upgrades need to be applied. The tag can be any name, but will normally refer to the ticket number that introduces the change.

Every new upgrade macro creates a new tagged version. A named version is simply a tagged version for which a copy of the relevant configuration metadata is made available.

Named versions for system releases are typically created at the end of the release process. The associated upgrade macro is typically only required in order to create the new name tag and, therefore, does not normally alter the application configuration.

Application configurations can reference the configuration metadata as follows:

# Refer to the HEAD version
# (typically you wouldn't do this since no upgrade process is possible)
# For flat metadata
# For hierarchical metadata

# Refer to a named or tagged version in the flat metadata
# Refer to a named or tagged version in the hierarchical metadata

If a version is defined then the Rose utilities will first look for the corresponding named version. If this cannot be found then the HEAD version is used and, if an upgrade is available, a warning will be given to indicate that the configuration metadata being used requires upgrade macros to be run. If the version defined does not correspond to a tagged version then a warning will be given.


If a hierarchical structure for the metadata is being used, the HEAD tag must be specified explicitly.

When to create named versions

One option is to create a new named version for each release of your system. This makes it easy for users to understand. However, if there is a new release which does not require a change to the metadata then you will still have to create a new copy and force the user to go through a null upgrade which may not be desirable. An alternative is to only create a new named version at releases which require changes. The name then indicates the metadata is relevant for a particular release and all subsequent releases (unless an upgrade macro is available to a later release).


It is also possible to make any tagged version between releases a named version, but it will usually be better not to. In which case, the user will be using HEAD and will be prompted to upgrade (which is probably what you want if you’re not using a release).

Sharing metadata between different executables

If two different commands share the majority of their inputs then you may choose to use the same configuration metadata for both commands. Any differences (in terms of available inputs) can then be triggered by the command in use. Whether this is desirable will partly depend on how many of the inputs are shared.

One downside of sharing metadata is that your application configuration may contain (ignored) settings which have no relevance to the command you are using.


We intend to introduce support for configuration metadata to include / inherit from other metadata. This may mean that it makes sense to have separate metadata for different commands even when the majority of inputs are shared.

Another reason you may want to share metadata is if you have two related commands which you want to configure using the same set of inputs (i.e. a single application configuration).

This works by setting an alternate command in the application configuration and then using the --command-key option to rose app-run.

Using development versions of upgrade macros

Users will be able to test out development versions of the upgrade macros by adding a working copy of the relevant branch into their metadata search path. However, care must be taken when doing this. Running the upgrade macro will change the rose-app.conf|meta setting to refer to the new tag. If the upgrade macro is subsequently changed or other upgrade macros are added to the chain prior to this tag (because they get committed to the trunk first) then this will result in application configurations which have not gone through the correct upgrade process. Therefore, when using development versions of the upgrade macros it is safest to not commit any resulting changes (or to use a branch of the suite which you are happy to discard).

Metadata Mini-Language

The Rose metadata mini-language supports writing a logical expression in Python-like syntax, using variable IDs to reference their associated values.

Expressions are set as the value of metadata properties such as rose-meta.conf[SETTING]fail-if and rose-meta.conf[SETTING]range.

The language is a small sub-set of Python - a limited set of operators is supported.


No built-in object methods, functions, or modules are supported - neither are control blocks such as if/for, statements such as del or with, or defining your own functions or classes. Anything that requires that kind of power should be in proper Python code as a macro.

Nevertheless, the language allows considerable power in terms of defining simple rules for variable values.


The following logical operators are supported:

and   # Logical AND
or    # Logical OR
not   # Logical NOT

The following numeric operators are supported:

+     # add
-     # subtract
*     # multiply
**    # power or exponent - e.g. 2 ** 3 implies 8
/     # divide
//    # integer divide (floor) - e.g. 3 // 2 implies 1
%     # remainder e.g. 5 % 3 implies 2

The following string operators are supported:

+      # concatenate - e.g. "foo" + "bar" implies "foobar"
*      # self-concatenate some number of times - e.g. "foo" * 2 implies "foofoo"
%      # formatting - e.g. "foo %s baz" % "bar" implies "foo bar baz"

# Where m, n are integers or expressions that evaluate to integers
# (negative numbers count from the end of the string):
[n]   # get nth character from string - e.g. "foo"[1] implies "o"
[m:n] # get slice of string from m to n - e.g. "foobar"[1:5] implies "ooba"
[m:]  # get slice of string from m onwards - e.g. "foobar"[1:] implies "oobar"
[:n]  # get slice of string up to n - e.g. "foobar"[:5] implies "fooba"

The following comparison operators are supported:

is    # Is the same object as (usually used for 'is none')
<     # Less than
>     # Greater than
==    # Equal to
>=    # Greater than or equal to
<=    # Less than or equal to
!=    # Not equal to

The following membership operator is supported:

in     # contained in (True/False)
# e.g. "oo" in "foobar" implies True; "foo" in ["bar", "baz"] implies False

# To test for lack of membership, combine with the "not" logical operator:
not in # not contained in, i.e. the opposite to that described above


The in operator functions as it does in Python Python in operator. Note the difference between testing if a setting’s value is equal to some possibility contained within a collection, for example as in:

place in ['here', 'there']  # i.e: place == 'here' or place == 'there'

and testing if it is a substring, as in:

place in 'there'  # place could be 'there', 'here', 'er', 'the', etc.


Operator precedence is intended to be the same as Python. However, with the current choice of language engine, the % and // operators may not obey this - make sure you force the correct behaviour using brackets.


The following are special constants:

None  # Python None
False # Python False
True  # Python True

Using Variable IDs

Putting a variable ID in the expression means that when the expression is evaluated, the string value of the variable is cast and substituted into the expression.

For example, if we have a configuration that looks like this:


and an expression in the configuration metadata:

namelist:zoo=elephant_mood != 'annoyed' and num_elephants >= 2

then the expression would become:

'peaceful' != 'annoyed' and 2 >= 2

If the variable is not currently available (ignored or missing) then the expression cannot be evaluated. If inter-variable comparisons are not allowed for the expression’s parent option (such as with rose-meta.conf[SETTING]trigger and rose-meta.conf[SETTING]range) then referencing other variable IDs is not allowed.

In this case the expression would be false.

You may use this as a placeholder for the current variable ID - for example, the fail-if expression:

fail-if=namelist:foo=bar > 100 and namelist:foo=bar % 2 == 1

is the same as:

fail-if=this > 100 and this % 2 == 1


The syntax has some special ways of dealing with variable values that are arrays - i.e. comma-separated lists.

You can refer to a single element of the value for a given variable ID (or this) by suffixing a number in round brackets - e.g.:


references the second element in the value for bar in the section namelist:foo. This follows Fortran index numbering and syntax, which starts at 1 rather than 0, i.e. foo(1) references the first element in the array foo.

If we had a configuration:

bar='a', 'b', 'c', 'd'

namelist:foo=bar(2) would get substituted in an expression with 'b' when the expression was evaluated. For example, an expression that contained:

namelist:foo=bar(2) == 'c'

would be evaluated as:

'b' == 'c'

Should you wish to make use of the array length in an expression you can make use of the len function, which behaves in the same manner as the Python len and Fortran size equivalents to return the array length. For example:

len(namelist:foo=bar) > 3

would be expanded to:

4 > 3

and evaluate as true.

There are two other special array functions, any and all, which behave in a similar fashion to their Python and Fortran equivalents, but have a different syntax.

They allow you to write a shorthand expression within an any() or all() bracket which implies a loop over each array element. For example:

any(namelist:foo=bar == 'e')

evaluates true if any elements in the value of bar in the section namelist:foo are 'e'. For the above configuration snippet, this would be expanded before evaluation to be:

'a' == 'e' or 'b' == 'e' or 'c' == 'e' or 'd' == 'e'


all(namelist:foo=bar == 'e')

evaluates true if all elements are 'e'. Again, with the above configuration, this would be expanded before proper evaluation:

'a' == 'e' and 'b' == 'e' and 'c' == 'e' and 'd' == 'e'


Rose uses an external engine to evaluate the raw language string after variable IDs and any any() and all() functions have been substituted and expanded.

The current choice of engine is Jinja2, which is responsible for the details of the supported Pythonic syntax. This may change.


Do not use any Jinja2-specific syntax.

Config Editor Ignored Mechanics

This describes the intended behaviour of rose config-edit when there is an ignored state mismatch for a setting - e.g. a setting might be enabled when it should be trigger-ignored.

rose config-edit differs from the strict command line macro equivalent (rose macro) because the Switch Off Metadata mode and accidentally metadata-less configurations need to be presented in a nice way without lots of unnecessary errors. rose config-edit should only report the errors where the state is definitely wrong or makes a material difference to the user.

The table contains error and fixing information for some varieties of ignored state mismatches. The actual situations are considerably more varied, given section-ignoring and latent variables - the table holds the most important varieties.

The State contains the actual states. The Trigger State column contains the trigger-mechanism’s expected states. The states can be:

IT - !!

trigger ignored

IU - !

user ignored

E - (normal)


A subset of possible ignored/enabled states, errors and fixes:


Trigger State


Display Error?

User Options























not trigger





not trigger




See [1]






See [2]

















not trigger





not trigger

























not trigger





not trigger