If there's the need to make different insert/update/replace queries, one would like to notify changes with notifyChanges only at the end of the whole process. As of now, changes are automatically notified at the end of each insert/update/delete. By overriding the generated ContentProvider one could be able to override the default implementation. Also, allowing the ContentProvider to be inherited from could let developers use some custom logic for join fake-Uris (which is something I'm doing) without having to manually edit the generated ContentProvider, by implementing that very logic in the inherited class.

Imagine we have a database with current database version 4. Let's add a table called "Tabs" on database version 5.

{
            "table_name":"Tabs",
            "version":5,
            "fields":[
                {
                    "name":"id",
                    "type":"long",
                    "is_primary_key":true,
                    "is_id":true
                        },
                {
                    "name":"name",
                    "type":"text"
                }
            ]
        }

The code generated is the following :

// Version 6 : Creation of the table
        public static void upgradeTable(SQLiteDatabase db, int oldVersion, int newVersion) {

            if (oldVersion < 1) {
                Log.i(LOG_TAG, "Upgrading from version " + oldVersion + " to " + newVersion 
                        + ", data will be lost!");

                db.execSQL("DROP TABLE IF EXISTS " + TABLE_NAME + ";");
                createTable(db);
                return;
            }


            if (oldVersion != newVersion) {
                throw new IllegalStateException("Error upgrading the database to version " 
                        + newVersion);
            }
        }

The SQLiteOpenHelper with its method getDatabaseLocked detects a new version and then launches "onUpgrade(db, version, mNewVersion);.

Tabs.upgradeTable(db, oldVersion, newVersion); is called, oldVersion=4 and newVersion=5 so the first condition is false and the second is true and we throw an exception. Shouldn't we add a mechanism to force the table creation?

Julien

I'm a newbie to SQLite. So please forgive. But I did receive this error:

Exception in thread "main" java.lang.IllegalArgumentException: Field "eventId" | is_autoincrement can only be used on an integer type field at com.foxykeep.cpcodegenerator.model.FieldData.(FieldData.java:44) at com.foxykeep.cpcodegenerator.model.TableData.(TableData.java:40) at com.foxykeep.cpcodegenerator.model.TableData.getClassesData(TableData.java:66) at com.foxykeep.cpcodegenerator.Main.main(Main.java:100)

I changed the code on line 44 to

if (dbIsAutoincrement && !type.equals("integer") && !type.equals("long")) { throw new IllegalArgumentException("Field "" + name + "" | is_autoincrement can " + "only be used on an integer type field"); }

My question is...is autoincrement valid only on int types? Is long OK?

I'd like the ability to add "enum" classes and refer to them in a field definition. And specify in the field definition to store a string or an int.

Could create something like:

"enums": [ { "enum_name": "ReadFlag", "enum_values": [ { "value": "UNREAD", "intvalue": "0", "string_value": "UNREAD" }, { "value": "READ", "intvalue": "1", "string_value": "READ" }, { "value": "UNSPECIFIED", "intvalue": "-1", "string_value": "UNSPECIFIED" }, ] },

and then the resulting class would be:

public enum ReadStatus {

UNREAD, READ, UNKNOWN;

/**
 * <p>
 * Returns the enumerated value for the specified integer. If the value is
 * out of range, UNKOWN is returned.
 * <ul>
 * <li>0 - {@code UNREAD}
 * <li>1 - {@code READ}
 * </ul>
 * 
 * @param in
 * @return the enumeration
 */
public static ReadStatus forInt(int in) {
    switch (in) {
    case 0:
        return UNREAD;
    case 1:
        return READ;
    }
    return UNKNOWN;
}

/**
 * /**
 * <p>
 * Returns the enumerated value for the specified integer. If the value is
 * out of range, UNKOWN is returned.
 * <ul>
 * <li>"UNREAD" - {@code UNREAD}
 * <li>"READ" - {@code READ}
 * </ul>
 * 
 * @param in
 * @return the enumeration
 */
public static ReadStatus forString(String in) {
    if ("READ".equalsIgnoreCase(in))
        return READ;
    if ("UNREAD".equalsIgnoreCase(in))
        return UNREAD;
    return UNKNOWN;
}

/**
 * Returns an integer representation of the enumeration
 * 
 * @return an integer representation of the enumeration
 */
public int toInt() {
    switch (this) {
    case UNREAD:
        return 0;
    case READ:
        return 1;
    default:
        return -1;
    }
}

/**
 * Returns an string representation of the enumeration
 * 
 * @return an string representation of the enumeration
 */
@Override
public String toString() {
    switch (this) {
    case UNREAD:
        return "UNREAD";
    case READ:
        return "READ";
    default:
        return null;
    }
}

}

Then a reference something like:

"tables": [ { "table_name": "TheTable", "fields": [ { "name": "status", "type": "int", "enumeration": "ReadFlag" },

The resulting table class would be:

public class TheTable { private ReadFlag status; }

With the appropriate SQL inserts doing a "toInt()" or "toString()" and the appropriate cursor reads doing the "Enumeration.forInt()" or "Enumeration.forString()" calls.

What do you think? Sorry, I'm just really fond of enumerations.

Was thinking of offering some code. But in the current Code Generator's construction, do you think this would be incredibly difficult or just a bit trivial?

Hi,

First of all thank you so much for this, it has been great, I really like using this together with dataDroid.

Now usually when you need to perform a query with a JOIN on your content provider you need to setup a new Uri on your URI matcher to "catch" the query correctly and route your code accordingly.

As of right now I'm unable to find support for this on this tool, so how would you suggest to do this?, maybe supporting this on the tool itself or changing the generated code afterwards.

Currently, the logging for table create/update is generated like so:

    if (ACTIVATE_ALL_LOGS) {
        Log.d(LOG_TAG, "MyTable | createTable start");
    }
    MyTable.createTable(db);
    if (ACTIVATE_ALL_LOGS) {
        Log.d(LOG_TAG, "MyTable | createTable end");
    }

If one has a number of tables, this begins to look very cluttered in the source. It might be better to have this logging pushed into the beginning and end of the createTable method (and the updateTable methods as well).

E.g.

public static void createTable(SQLiteDatabase db) {
    if (ACTIVATE_ALL_LOGS) {
        Log.d(LOG_TAG, "MyTable | createTable start");
    }
    db.execSQL("CREATE TABLE " + TABLE_NAME + " (" +
                    Columns.ID.getName() + " " + Columns.ID.getType()+ ", " +
                    Columns.NAME.getName() + " " + Columns.NAME.getType() + ", " +
    ");");
    if (ACTIVATE_ALL_LOGS) {
        Log.d(LOG_TAG, "MyTable | createTable end");
    }
}

Hi,

When we do a request, the notificationUri is not the given Uri but the authority Uri. It causes a cursor to be notified each time there is a change somewhere in the database. The notificationUri should be the queried Uri.

Current source generated :

Uri notificationUri = LeGorafiContent.CONTENT_URI; [...] if ((c != null) && !isTemporary()) { c.setNotificationUri(getContext().getContentResolver(), notificationUri); }

Julien

When copy the sample.json to input using "cp ../input/example/sample.json ./input", it will copy the json as a file but not in input directory as there is no "input" folder. Fix it by create "input" directory explicitly before copying the json sample.

Is there any chance you will add an autoincrement option for the primary key generation? It would be very useful when handling entities without a unique identificator, as it would generate it automatically.

        static String getBulkInsertString() {
            return new StringBuilder("INSERT INTO ")
                                     .append(TABLE_NAME).append(" ( ").append(",")
                                     .append(Columns.X.getName()).append(", ")
                                     .append(Columns.Y.getName())
                                     .append(" ) VALUES (, ?, ?)").toString();
        }

which results in

android.database.sqlite.SQLiteException: near ",": syntax error: , while compiling: INSERT INTO routePoint ( , lat, lon ) VALUES (, ?, ?)

It should probably produce:

        static String getBulkInsertString() {
            return new StringBuilder("INSERT INTO ")
                                     .append(TABLE_NAME).append(" ( ")
                                     .append(Columns.X.getName()).append(", ")
                                     .append(Columns.Y.getName())
                                     .append(" ) VALUES (?, ?)").toString();
        }

The generated provider has the whereWithId method that is used by single-entities queries such as query, update, delete. If a huge number of queries is executed on slow devices (I am using a Nexus One for testing purposes) a warning such as the following can be issued:

Reached MAX size for compiled-sql statement cache for database <nameofdatabase>; i.e., 
NO space for this sql statement in cache: 
SELECT COLUMN_NAME FROM TABLE_NAME WHERE _ID = '1337'. 
Please change your sql statements to use '?' for bindargs, instead of using actual values

This can be solved by using the _ID column as a parameter. I have fixed this in my fork, I will make a pull request asap.

All that will be needed is to defer constant init until the content provider's creation, so we have a context to get the authority from a string resource

This will be useful when you need to define a content provider for a debug or for a release build. or any build variant if you are using gradle

First, I'd like to say great work.

I have a suggestion:

I think you should change the insert function from

id = db.insert(uriType.getTableName(), "foo", values);

to

id = db.replace(uriType.getTableName(), "foo", values);

This means, as a developer, you no longer need to query the table if the record exists or not to decide weather to do an update or an insert.

This works great with DataDroid bulk operations

Thanks

Hi,

First of all, thanks for this awesome tool ;) I think we should do a null check in bindValuesInBulkInsert for every DB type, not only String. Because as stated in the documentation for getAsLong for instance:

Returns the Long value, or null if the value is missing or cannot be converted

In my case, I do not set a value in every column when inserting new rows using a bulkInsert and it yield a NPE. I think there will be a special case for the _id column. If we haven't set a value for _id we shouldn't set a default one and instead let the system auto_increment this column. Something like:

Long longValue = values.getAsLong(Columns.ID.getName());
if(longValue != null)
    stmt.bindLong(i++, longValue);
else
    i++;

Or, more compact

Long longValue = values.getAsLong(Columns.ID.getName());
if(longValue != null)
    stmt.bindLong(i, longValue);
i++;

What do you think?