I am trying to model a type as a union where each member of that union has properties in common with all other members.
I am currently achieving this like so:
type alias File = {
name : String
}
type CommonFileState extra = CommonFileState {
id : String
, file : File
} extra
type alias ValidFileState = CommonFileState {
validatedAt : Int
}
type alias InvalidFileState = CommonFileState {
reason : String
}
type alias LoadingFileState = CommonFileState {}
type FileState = Valid ValidFileState | Invalid InvalidFileState | Loading LoadingFileState
Now if I want to read one of those common properties on any given FileState, I must match against each member of the union:
getId : FileState -> String
getId fileState = case fileState of
Valid (CommonFileState {id} extra) -> id
Invalid (CommonFileState {id} extra) -> id
Loading (CommonFileState {id} extra) -> id
This feels wrong to me, because I have to duplicate the property access for each member. If I needed to manipulate this property somehow (e.g. concatenating something onto the string), I would also have to duplicate this.
I want to be able to easily access common properties of my union, and operate on those common properties.
When I started searching for other ways to do this, I found one alternative was to nest the union inside a record, which also holds the common properties:
type alias ValidCurrentFileState = {
validatedAt : Int
}
type alias InvalidCurrentFileState = {
reason : String
}
type alias LoadingCurrentFileState = {}
type CurrentFileState = Valid ValidCurrentFileState | Invalid InvalidCurrentFileState| Loading LoadingCurrentFileState
type alias File = {
name : String
}
type alias FileState = {
id : String
, file : File
, currentState : CurrentFileState
}
getId : FileState -> String
getId {id} = id
However this is awkward because I have to name the nested union, which adds a level of unnecessary indirection: "file state" and "current file state" are conceptually the same.
Are there any other ways of doing this which don't have the problems I mentioned?
I think you are thinking about this the wrong way around.
The purpose of modelling (in Elm) is capture the possible states of your data, and to exclude - in your model - 'impossible' states, so that the compiler can statically prevent the code every creating such states.
Once you're happy with your model, you write the helpers you need to make your core logic easy to express and to maintain.
I suspect I would normally go with your second approach, but I don't know all the issues you need to account for.
Related
I'm trying to implement a dynamic form in Elm 0.19 using hecrj/composable-form.
I receive a json with the fields, their descriptions, etc, so I don't know beforehand how many fields it will have.
So the traditional way of defining a form:
Form.succeed OutputValues
|> Form.append field1
|> Form.append field2
doesn't work because I don't know the OutputValues structure beforehand.
I've seen there is a function Form.list which looks like a promising path, though it seems to expect all fields equal, which is not my case, I may have a text field and a select field for example.
Is there any straight forward way of doing this with this library?
Thank you.
The form library doesn't explicitly support what you're trying to do, but we can make it work!
tldr;
Here's my example of how you can take JSON and create a form: https://ellie-app.com/bJqNh29qnsva1
How to get there
Form.list is definitely the promising path. You're also exactly right that Form.list requires all of the fields to be of the same type. So let's start there! We can make one data structure that can hold them by making a custom type. In my example, I called it DynamicFormFieldValue. We'll make a variant for each kind of field. I created ones for text, integer, and select list. Each one will need to hold the value of the field and all of the extras (like title and default value) to make it show up nicely. This will be what we decode the JSON into, what the form value is, and what the form output will be. The resulting types looks like this:
type alias TextFieldRequirements =
{ name : String
, default : Maybe String
}
type alias IntFieldRequirements =
{ name : String
, default : Maybe Int
}
type alias SelectFieldRequirements =
{ name : String
, default : Maybe String
, options : List ( String, String )
}
type DynamicFormFieldValue
= TextField String TextFieldRequirements
| IntField Int IntFieldRequirements
| SelectField String SelectFieldRequirements
To display the form, you just need a function that can take the form value and display the appropriate form widget. The form library provides Form.meta to change the form based on the value. So, we will pattern match on the custom type and return Form.textField, Form.numberField, or Form.selectField. Something like this:
dynamicFormField : Int -> Form DynamicFormFieldValue DynamicFormFieldValue
dynamicFormField fieldPosition =
Form.meta
(\field ->
case field of
TextField textValue ({ name } as requirements) ->
Form.textField
{ parser = \_ -> Ok field
, value = \_ -> textValue
, update = \value oldValue -> TextField value requirements
, error = always Nothing
, attributes =
{ label = name
, placeholder = ""
}
}
IntField intValue ({ name } as requirements) ->
Form.numberField
{ parser = \_ -> Ok field
, value = \_ -> String.fromInt intValue
, update = \value oldValue -> IntField (Maybe.withDefault intValue (String.toInt value)) requirements
, error = always Nothing
, attributes =
{ label = name
, placeholder = ""
, step = Nothing
, min = Nothing
, max = Nothing
}
}
SelectField selectValue ({ name, options } as requirements) ->
Form.selectField
{ parser = \_ -> Ok field
, value = \_ -> selectValue
, update = \value oldValue -> SelectField value requirements
, error = always Nothing
, attributes =
{ label = name
, placeholder = ""
, options = options
}
}
)
Hooking this display function up is a bit awkward with the library. Form.list wasn't designed with use-case in mind. We want the list to stay the same length and just be iterated over. To achieve this, we will remove the "add" and "delete" buttons and be forced to provide a dummy default value (which will never get used).
dynamicForm : Form (List DynamicFormFieldValue) (List DynamicFormFieldValue)
dynamicForm =
Form.list
{ default =
-- This will never get used
TextField "" { name = "", default = Nothing }
, value = \value -> value
, update = \value oldValue -> value
, attributes =
{ label = "Dynamic Field Example"
, add = Nothing
, delete = Nothing
}
}
dynamicFormField
Hopefully the ellie example demonstrates the rest and you can adapt it to your needs!
I'm trying to extract some information from some records using extensible records. If I create a function taking the extensible record type and returning a string and use that within a case statement then there are no issues (namedToString, in the below example). However, if I attempt to use a function passed as a parameter (stringFromNamed) I will get an error complaining:
This business value is a:
Business
But stringFromNamed needs the 1st argument to be:
Named a -> String (edited)
from the example code below:
type alias Named a =
{ a | name : String }
type alias Person =
Named { address : String }
type alias Business =
Named { employeeCount : Int }
type Change
= PersonUpdate Person
| BusinessUpdate Business
namedToString : Named a -> String
namedToString changeFields =
changeFields.name
changeToString : (Named a -> String) -> Change -> String
changeToString stringFromNamed change =
case change of
PersonUpdate person ->
-- This works
namedToString person
BusinessUpdate business ->
-- This will cause the error
stringFromNamed business
This example includes just the required code but a more complete example can be found at https://ellie-app.com/77nCPLh55j3a1
What is causing the issue and how can I achieve my goal of passing in a function which will extract some information from an extensible record?
I discovered this issue in the Elm compiler repo: https://github.com/elm/compiler/issues/1959
This appears to be a bug in the compiler which can be worked around if you just want to pass an external record of one type to a function parameter. You can do this by removing the type signature of the higher order function, changeToString in the example above.
Unfortunately, if we wanted to perform the same action from the function parameter, e.g. stringFromNamed, on every case the error will return so another workaround must be found.
The workaround I'm using is to create a new record type which includes exactly the fields from the extensible record and then creating an instance of this from the fields of the other records adhering to the extensible record type. Not a big problem with only a couple of cases and an extensible record with only one field but this doesn't scale especially well. Example below:
type alias Named a =
{ a | name : String }
type alias OnlyNamed =
Named {}
type alias Person =
Named { address : String }
type alias Business =
Named { employeeCount : Int }
type Change
= PersonUpdate Person
| BusinessUpdate Business
type Msg
= UpdateCurrentChange Change
namedToString : Named a -> String
namedToString changeFields =
changeFields.name
changeToString : (OnlyNamed -> String) -> Change -> String
changeToString stringFromNamed change =
case change of
PersonUpdate { name } ->
stringFromNamed { name = name }
BusinessUpdate { name } ->
stringFromNamed { name = name }
I think the issue is that the function you're passing in takes Named a, and Elm doesn't know what the a refers to, so it ends up deciding it's a type mismatch. If you change the type annotation to (Named Business -> String) -> Change -> String it will compile and work.
I am a super elm begginer and trying to make app.
Currently I am struggling to make landing page and http request to a server.
But, I am stuck here...
I have init function something like this below.
init : () -> Url.Url -> Nav.Key -> ( Model, Cmd Msg )
init flags url key =
Model key TopPage
|> goTo (Route.parse url)
The definition of my Model is below.
-- MODEL
type alias Model =
{ key : Nav.Key
, page : Page
, name : String
, tags : List Tag
, jwt : String }
and, goTo function is below.
goTo : Maybe Route -> Model -> ( Model, Cmd Msg )
goTo maybeRoute model =
case maybeRoute of
Nothing ->
( { model | page = NotFound }, Cmd.none )
Just Route.Top ->
( { model | page = TopPage }, Cmd.none )
...
type Route is below.
type Route
= Top
| User String
| Repo String String
parse : Url -> Maybe Route
parse url =
Url.Parser.parse parser url
parser : Parser (Route -> a) a
parser =
oneOf
[ map Top top
, map User string
, map Repo (string </> string)
]
but following error has occured.
-- TYPE MISMATCH -------------------------------------------------- src/Main.elm
This function cannot handle the argument sent through the (|>) pipe:
54| Model key TopPage
55| |> goTo (Route.parse url)
^^^^^^^^^^^^^^^^^^^^^
The argument is:
String -> List Tag -> String -> Model
But (|>) is piping it a function that expects:
Model
What did I make mistake here?....
Your Model type has five fields, but in the line
Model key TopPage
you are only providing values for the first two of the five. You are missing values for the name, tags and jwt fields. Provide values for these and the problem should go away.
When you declare a type alias such as Model, Elm creates a constructor function also named Model. Elm functions support partial application, in that if you pass in values for some but not all of the arguments, you end up with a function that takes in the rest of the values. You provided two arguments, so you end up with a function that takes three arguments and returns a Model.
There are two ways of building a value of a type. Given a simple example of a Person type alias:
type alias Person = { name : String, age : Int }
You can construct a value by specifying all fields (note that you don't have to specify Person in the constructor; Elm's compiler is smart enough to know it by its shape):
jane : Person
jane = { name = "Jane", age = 35 }
Or you can build a value by using the type name and specify each field's values in the order in which they were defined. In this style, you can think of Person acting like a function with two parameters that returns a Person value.
jane : Person
jane = Person "Jane" 35
In each case, you have to specify all fields of the type when you construct it in order to obtain a complete Person value. However, that is not the complete story. It is possible to leave off the age parameter when constructing a Person, but the result isn't a Person, it's a function that takes an age and returns a Person. In other words,
janeAged : Int -> Person
janeAged = Person "Jane"
You can strip off as many parameters from the end as you'd like to make more variations on that constructor, even stripping out all parameters:
somebody : String -> Int -> Person
somebody = Person
Back to your example. You are constructing a Model value by only specifying two parameters (Model key TopPage). The value of that expression does not result in a Model, but in a function that takes three more parameters to create a Model. And that's why the error message indicated you need three parameters to construct a model.
You need to specify all values of Model when creating it.
I want to add a certain value to a list. Both are inside my ELM-model:
type alias Model =
{ syllables : List Syllable
, words : List Word
, newSyllable : String
, newWord : String
}
I want to add the newSyllable value to the list of syllables, when I click the button.
I placed this attribute inside my view:
onClick TransferSyllable
Everything works right, but I wonder how I can transfer one value of my model into the list of values!?
Thanks.
EDIT:
This is my definition of "Syllable":
type alias Syllable =
{ content : String
, start : Bool
, mid : Bool
, end : Bool
}
I want to insert the value to the end of the list.
Since Syllable is a record type with four fields, and newSyllable is just a string, you'll need a function that turns a String into a Syllable. I'll assume that function has the signature:
makeSyllable : String -> Syllable
Adding the syllable onto the end of the list can be done using List.append. Since append takes a List a, you'll need to add brackets around newSyllable when passing it to append:
{ model | syllables = List.append model.syllables [ makeSyllable model.newSyllable ] }
When running a particular unit-test, I am getting the exception:
Caused by: org.hibernate.HibernateException: Errors in named queries: UPDATE_NEXT_FIRE_TIME
at org.hibernate.impl.SessionFactoryImpl.<init>(SessionFactoryImpl.java:437)
at org.hibernate.cfg.Configuration.buildSessionFactory(Configuration.java:1385)
at org.hibernate.cfg.AnnotationConfiguration.buildSessionFactory(AnnotationConfiguration.java:954)
at org.hibernate.ejb.Ejb3Configuration.buildEntityManagerFactory(Ejb3Configuration.java:891)
... 44 more
for the named query defined here:
#Entity(name="fireTime")
#Table(name="qrtz_triggers")
#NamedQueries({
#NamedQuery(
name="UPDATE_NEXT_FIRE_TIME",
query= "update fireTime t set t.next_fire_time = :epochTime where t.trigger_name = 'CalculationTrigger'")
})
public class JpaFireTimeUpdaterImpl implements FireTimeUpdater {
#Id
#Column(name="next_fire_time", insertable=true, updatable=true)
private long epochTime;
public JpaFireTimeUpdaterImpl() {}
public JpaFireTimeUpdaterImpl(final long epochTime) {
this.epochTime = epochTime;
}
#Override
public long getEpochTime() {
return this.epochTime;
}
public void setEpochTime(final long epochTime) {
this.epochTime = epochTime;
}
}
After debugging as deep as I could, I've found that the exception occurs in w.statement(hqlAst) in QueryTranslatorImpl:
private HqlSqlWalker analyze(HqlParser parser, String collectionRole) throws QueryException, RecognitionException {
HqlSqlWalker w = new HqlSqlWalker( this, factory, parser, tokenReplacements, collectionRole );
AST hqlAst = parser.getAST();
// Transform the tree.
w.statement( hqlAst );
if ( AST_LOG.isDebugEnabled() ) {
ASTPrinter printer = new ASTPrinter( SqlTokenTypes.class );
AST_LOG.debug( printer.showAsString( w.getAST(), "--- SQL AST ---" ) );
}
w.getParseErrorHandler().throwQueryException();
return w;
}
Is there something wrong with my query or annotations?
NamedQuery should be written with JPQL, but query seems to mix both names of persistent attributes and names of database columns. Names of database columns cannot be used in JPQL.
In this case instead of next_fire_time name of the persistent attribute epochTime should be used. Also trigger_name looks more like name of the database column than name of the persistent attribute, but it seems not to be mapped in your current class at all. After it is mapped, query is as follows:
update fireTime t set t.epochTime = :epochTime
where t.triggerName = 'CalculationTrigger'
If SQL query is preferred, then #NamedNativeQuery should be used instead.
As a side note, JPA 2.0 specification doesn't encourage changing primary key:
The application must not change the value of the primary key[10]. The
behavior is undefined if this occurs.[11]
In general entities are not aware of changed made via JPQL queries. That gets especially interesting when trying to refresh entity that does not exist anymore (because primary key was changed).
Additionally naming is little bit confusing:
Name of the class looks more like name of the service class
than name of the entity.
Starting name of the entity with lower
case letter is rather rare style.
Name of the entity, name of the
table and name of the class do not match too well.