fix broken docs formatting and provide uniform method for multiple names

This commit is contained in:
Jonathan Shook 2024-07-26 16:57:40 -05:00
parent 2d30d69f1f
commit a06b62c12d

View File

@ -70,7 +70,7 @@ import java.util.function.LongFunction;
* specification tests under the workload_definition folder of the adapters-api module.
* </P>
*
*
* <HR/>
* <H2>Op Template Parsing</H2>
* <OL>
* <LI>Rule #1: All op templates are parsed into an internal normalized structure which contains:
@ -92,36 +92,41 @@ import java.util.function.LongFunction;
* The net effect of these rules is that the NoSQLBench driver developer may safely use functional forms to access data
* in the op template, or may decide that certain op fields must only be provided in a static way per operation.
* </P>
*
* <HR/>
* <H2>Distinguishing Op Payload from Op Config</H2>
* <P>When a user specifies an op template, they may choose to provide only a single set of op fields without
* distinguishing between config or payload, or they may choose to directly configure each.
* <H4>Example:</H4>
*
* <PRE>{@code
* ops:
* # both op and params explicitly named
* op1:
* op:
* opfield1: value1
* params:
* param2: value2
* # neither op field nor params named, so all assumed to be op fields
* op2:
* opfield1: value1
* param2: value2
* # in this case, if param2 is meant to be config level,
* # it is a likely config error that should be thrown to the user
* # only op field explicitly named, so remainder automatically pushed into params
* op3:
* op:
* opfield1: value1
* param2: value2
* # only params explicitly named, so remainder pushed into op payload
* op4:
* params:
* param2: value2
* opfield1: value1
* }</PRE>
*
* # both op and params explicitly named
* op1:
* op:
* opfield1: value1
* params:
* param2: value2
*
* # neither op field nor params named, so all assumed to be op fields
* op2:
* opfield1: value1
* param2: value2
*
* # in this case, if param2 is meant to be config level,
* # it is a likely config error that should be thrown to the user
* # only op field explicitly named, so remainder automatically pushed into params
* op3:
* op:
* opfield1: value1
* param2: value2
*
* # only params explicitly named, so remainder pushed into op payload
* op4:
* params:
* param2: value2
* opfield1: value1
* }</PRE></P>
*
* <p>
* All of these are considered valid constructions, and all of them may actually achieve the same result.
* This looks like an undesirable problem, but it serves to simplify things for users in one specific way: It allows
@ -135,37 +140,37 @@ import java.util.function.LongFunction;
* <HR></HR>
* <H2>Design Invariants</H2>
*
* <P>The above rules imply invariants, which are made explicit here. {@link ParsedOp}.</P>
* <P>The above rules imply invariants, which are made explicit here.</P>
*
* <P><UL>
* <H4>Single Purpose Fields</H4>
* <P><EM>You may not use an op field name or parameter name for more than one purpose.</EM></P>
*
* <LI><EM>You may not use an op field name or parameter name for more than one purpose.</EM>
* <UL>
* <LI>Treat all parameters supported by a driver adapter and it's op fields as a globally shared namespace, even if it
* is not.
* This avoids creating any confusion about what a parameter can be used for and how to use it for the right thing in
* the right place.
* For example, you may not use the parameter name `socket` in an op template to mean one thing and then use it
* <H4>Shared Namespace</H4>
* <P><EM>Treat all parameters supported by a driver adapter and it's op fields as a globally shared namespace, even
* if it is not.</EM></P>
*
* <P>This avoids creating any confusion about what a parameter can be used for and how to use it for the right thing in
* the right place. For example, you may not use the parameter name `socket` in an op template to mean one thing and then use it
* at the driver adapter level to mean something different. However, if the meaning is congruent, a driver developer
* may choose to support some cross-cutting parameters at the activity level. These allowances are explicit,
* however, as each driver dictates what it will allow as activity parameters.
* </LI>
* </UL>
* </LI>
* however, as each driver dictates what it will allow as activity parameters.</P>
*
* <H3>Layered Resolution</H3>
*
* <P><EM>Users may specify op payload fields within op params or activity params as fallback config sources in that
* order.</EM></P>
*
* <LI><EM>Users may specify op payload fields within op params or activity params as fallback config sources in that
* order.</EM>
* <UL>
* <LI>IF a name is valid as an op field, it must also be valid as such when specified in op params.</LI>
* <LI>If a name is valid as an op field, it must also be valid as such when specified in activity params, within the
* scope of {@link ParsedOp}</LI>
* <LI>When an op field is found via op params or activity params, it may NOT be dynamic. If dynamic values are intended
* to be provided
* at a common layer in the workload, then bindings support this already.</LI>
* to be provided at a common layer in the workload, then bindings or template variables support this already.</LI>
* </UL>
* </LI>
*
* <LI><EM>You must access non-payload params via Config-oriented methods.</EM>
* <H3>Configuration Fields</H3>
* <EM>You must access non-payload params via Config-oriented methods.</EM>
* <UL>
* <LI>Op Templates contain op payload data and op configs (params, activity params).</LI>
* <LI>You must use only {@link ParsedOp} getters with "...Config..." names, such as
@ -176,56 +181,49 @@ import java.util.function.LongFunction;
* </UL>
* </LI>
*
* <LI><EM>The user must be warned when a required or optional config value is missing from op params (or activity
* params), but a value
* of the same name is found in op payload fields.</EM>
* <UL>
* <LI>If rule #1 is followed, and names are unambiguous across the driver, then it is almost certainly a configuration
* error.</LI>
* </UL>
* </LI>
* <H3>Sanity Checks</H3>
*
* <LI><EM>When both an op payload field and a param field of the same name are defined through cascading configuration
* of param fields,
* the local op payload field takes precedence.</EM>
* <P><EM>The user must be warned when a required or optional config value is missing from op params (or activity
* params), but a value of the same name is found in op payload fields.</EM>
* If rule #1 is followed, and names are unambiguous across the driver, then it is almost certainly a configuration
* error.</P>
*
* <H3>Precedence</H3>
* <P>The order of precedence for values is:
* <UL>
* <LI>This is an extension of the param override rules which say that the closest (most local) value to an operation is
* the one that takes precedence.</LI>
* <LI>In practice, there will be no conflicts between direct static and dynamic fields, but there will be possibly
* between
* static or dynamic fields and parameters and activity params. If a user wants to promote an activity param as an
* override to existing op fields,
* template variables allow for this to happen gracefully. Otherwise, the order of precedence is 1) op fields 2) op
* params 3) activity params.</LI>
* </UL>
* </LI>
* <LI>op payload fields</LI>
* <LI>op param fields</LI>
* <LI>activity params (when enabled, see below)</LI>
* </UL>
* </P>
*
* <H3>Enabling Activity Params</H3>
* <P>If a user wants to allow an activity param as an default for an fields, they must publish the op field
* name in the configuration model for the activity. Otherwise it is an error to specify the value at the activity
* level.</P>
*
* <HR></HR>
*
* <H2>Op Payload Forms</H2>
* Field values can come from multiple sources. These forms and any of their combinations are supported.
*
* <H3>Static Op Fields</H3>
* <H4>Example:</H4>
* <PRE>{@code
* op:
* field1: value1
* field2:
* map3:
* key4: value4
* map5:
* key6: value6
* field7: false
* field8: 8.8
* field1: value1
* field2:
* map3:
* key4: value4
* map5:
* key6: value6
* field7: false
* field8: 8.8
* }</PRE>
* <p>
* As shown, any literal value of any valid YAML type, including structured values like lists or maps are accepted as
* static op template values. A static value is any value which contains zero bind points at any level.
*
* <H3>Dynamic Op Fields with Binding References</H3>
* <H4>Example:</H4>
* <PRE>{@code
* op:
* field1: "{binding1}"
@ -247,34 +245,32 @@ import java.util.function.LongFunction;
* null values invalid for ANY op template value.</P>
*
* <H3>Dynamic Op Fields with Binding Definitions</H3>
* <H4>Example:</H4>
*
* <PRE>{@code
* op:
* field1: "{{NumberNameToString()}}"
* field2: "value is: {{NumberNameToString()}}"
* field1: "{{NumberNameToString()}}"
* field2: "value is: {{NumberNameToString()}}"
* }
* </PRE>
* <p>
* This form has exactly the same effect as the previous example as long as your bindings definitions included:
* <PRE>{@code
* bindings:
* binding1: NumberNameToString();
* binding1: NumberNameToString();
* }</PRE>
*
* <H3>Dynamic Op Fields with Structure</H3>
* <H4>Example:</H4>
*
* <PRE>{@code
* field1:
* k1: "{binding1}
* k2: "literal value"
* field2:
* - "value3"
* - "{binding4}"
* - "a value: {binding5}"
* - "{{NumberNameToString}}"
* - "a value: {{NumberNameToString()}}"
* op:
* field1:
* k1: "{binding1}
* k2: "literal value"
* field2:
* - "value3"
* - "{binding4}"
* - "a value: {binding5}"
* - "{{NumberNameToString}}"
* - "a value: {{NumberNameToString()}}"
* }</PRE>
*
* <P>This example combines the previous ones with structure and dynamic values. Both field1 and field2 are dynamic,
@ -286,15 +282,14 @@ import java.util.function.LongFunction;
* configure your binding definitions thusly.</P>
*
* <H3>Op Template Params</H3>
* <H4>Example:</H4>
* <PRE>{@code
* params:
* prepared: true
* prepared: true
* ops:
* op1:
* field1: value1
* params:
* prepared: false
* op1:
* field1: value1
* params:
* prepared: false
* }</PRE>
* <p>
* The params section are the first layer of external configuration values that an op template can use to distinguish
@ -308,7 +303,6 @@ import java.util.function.LongFunction;
* down to each block and then down to each statement.
*
* <H3>Activity Params</H3>
* <H4>Example:</H4>
* <PRE>{@code
* ./nb run driver=... workload=... cl=LOCAL_QUORUM
* }</PRE>
@ -317,6 +311,19 @@ import java.util.function.LongFunction;
* another fallback source for configuration parameters. The {@link ParsedOp} implementation will automatically look
* in the activity parameters if needed to find a missing configuration parameter, but this will only work if
* the specific named parameter is allowed at the activity level.</P>
*
* <HR/>
* <H2>Alternate Names</H2>
* <P>Sometimes you may need to support more than one name for the same purpose. In such cases, there are helper
* methods which can be used to reduce from a set of possible field names to a single one.
* <UL>
* <LI>{@link #requiredFieldOf(String...)} and {@link #requiredFieldOf(List)} will find exactly one field within the set of
* possible fields.
* Zero or more than one will throw an error.</LI>
* <LI>{@link #optionalFieldOf(String...)} and {@link #optionalFieldOf(List)} will find exactly zero or one
* field within the set of possible fields. More than one will throw an error.</LI>
* </UL>
* </P>
*/
public class ParsedOp extends NBBaseComponent implements LongFunction<Map<String, ?>>, NBComponent, StaticFieldReader, DynamicFieldReader {
@ -571,15 +578,6 @@ public class ParsedOp extends NBBaseComponent implements LongFunction<Map<String
}
public <T> Optional<T> getOptionalStaticValue(List<String> fieldNames, Class<T> classOfT) {
for (String field : fieldNames) {
if (isStatic(field)) {
return Optional.ofNullable(tmap.getStaticValue(field, classOfT));
}
}
return Optional.empty();
}
public <T> Optional<T> takeOptionalStaticValue(String field, Class<T> classOfT) {
return Optional.ofNullable(tmap.takeStaticValue(field, classOfT));
}
@ -919,24 +917,6 @@ public class ParsedOp extends NBBaseComponent implements LongFunction<Map<String
return lfa;
}
public <FA, FE> LongFunction<FA> enhanceFunc(
LongFunction<FA> func,
List<String> fields,
Class<FE> type,
BiFunction<FA, FE, FA> combiner
) {
for (String field : fields) {
if (isDefined(field)) {
LongFunction<FE> fieldEnhancerFunc = getAsRequiredFunction(field, type);
LongFunction<FA> lfa = l -> combiner.apply(func.apply(l), fieldEnhancerFunc.apply(l));
return lfa;
}
}
throw new OpConfigError("One of the op fields from " + fields + " is required, but none was found " +
"in the op template named '" + this.getName() + "'");
}
/**
* <p>Enhance a {@link Function} with a named optional function IFF it exists.</p>
*
@ -969,23 +949,6 @@ public class ParsedOp extends NBBaseComponent implements LongFunction<Map<String
return lfa;
}
public <FA, FE> LongFunction<FA> enhanceFuncOptionally(
LongFunction<FA> func,
List<String> fields,
Class<FE> type,
BiFunction<FA, FE, FA> combiner
) {
for (String field : fields) {
Optional<LongFunction<FE>> fieldEnhancerFunc = getAsOptionalFunction(field, type);
if (fieldEnhancerFunc.isPresent()) {
LongFunction<FE> feLongFunction = fieldEnhancerFunc.get();
LongFunction<FA> lfa = l -> combiner.apply(func.apply(l), feLongFunction.apply(l));
return lfa;
}
}
return func;
}
public static enum SubOpNaming {
SubKey,
@ -1241,4 +1204,38 @@ public class ParsedOp extends NBBaseComponent implements LongFunction<Map<String
}
private String oneFieldOrNull(List<String> fields) {
int count = 0;
String found = null;
for (String field : fields) {
if (isDefined(field)) {
if (found == null) {
found = field;
} else {
throw new OpConfigError(STR."exactly one of \{fields} was required, but it was found as both \{found} and \{field}");
}
}
}
return found;
}
public Optional<String> optionalFieldOf(List<String> fields) {
return Optional.ofNullable(oneFieldOrNull(fields));
}
public Optional<String> optionalFieldOf(String... fields) {
return Optional.ofNullable(oneFieldOrNull(List.of(fields)));
}
public String requiredFieldOf(String... fields) {
return requiredFieldOf(List.of(fields));
}
public String requiredFieldOf(List<String> fields) {
String field = oneFieldOrNull(fields);
if (field==null) {
throw new OpConfigError(STR."no fields were found with name in (\{fields}) for op template \{this}");
}
return field;
}
}