移动端orm框架性能测评
flutter_orm_plugin 发布以来，不少团队试用了，我发现大家对这类数据库相关的库，第一反应就是性能如何，之前确实没做太多行业对比，最近觉得还是有必要做一下性能测试，给大家一个交代的。
在ios端，业界比较常用的orm框架应该是苹果官方推出的coredata，还有就是realm了。在android端orm框架我挑了三个比较常用的，greendao，realm和activeandroid。我会用flutter_orm_plugin跟上面提到的ios和android端orm框架做对比。 下面我会分别给出测试用例，测试代码，还有最终数据比较的结果。
测试用例
测试用例我列了以下这些
- 10000次插入数据
- 使用批量接口10000次插入数据
- 10000次读取数据
- 10000次修改数据
- 使用批量接口10000次修改数据
- 10000次删除数据
- 使用批量接口10000次删除数据
为什么会有普通插入数据和使用批量接口插入数据的区别，大部分orm框架都会对批量操作有一定的优化，所以需要对批量操作进行测试，但是在平时使用，不一定都能用上批量接口（例如多次数据操作不在同一代码块，或者在不同的模块中都要操作数据），所以我们会分别对普通操作和批量操作进行测试。
android 测试代码
首先我们给出flutter_orm_plugin 的测试代码，由于不想因为flutter和原生channel通讯产生误差，我们直接用Luakit来写lua代码做测试（greendao、realm、activeandroid、coredata都不涉及flutter和原生channel通讯），flutter_orm_plugin其实底层就是luakit的orm框架，这个不影响测试准确性。
循环插入
Luakit定义orm模型结构并做10000次插入，下面的代码是ios和android通用的。
local Student = { __dbname__ = "test.db", __tablename__ = "Student", studentId = {"TextField",{primary_key = true}}, name = {"TextField",{}}, claName = {"TextField",{}}, teacherName = {"TextField",{}}, score = {"RealField",{}}, } local params = { name = "Student", args = Student, } Table.addTableInfo(params,function () local studentTable = Table("Student”) for i=1,10000 do local s = { studentId = "studentId"..i, name = "name"..i, claName = "claName"..i, teacherName = "teacherName"..i, score = 90, } studentTable(s):save() end end)activeandroid定义orm模型结构并做10000次插入
@Table(name = "Students")public class Student extends Base { @Column(name = "studentId") public String studentId; @Column(name = "name") public String name; @Column(name = "claName") public String claName; @Column(name = "teacherName") public String teacherName; @Column(name = "score") public float score; public Student() { super(); } @Override public String toString() { return this.studentId; }}for (int i=0 ; i<10000 ;i++) { ActiveAndroid.beginTransaction(); Student s = new Student(); s.studentId = "studentId"+i; s.name = "name"+i; s.teacherName = "teacherName"+i; s.claName = "claName"+i; s.score = 90; s.save(); ActiveAndroid.setTransactionSuccessful(); ActiveAndroid.endTransaction();}realm android 定义orm模型结构并做10000次插入
public class StudentRealm extends RealmObject { @PrimaryKey private String studentId; @Required private String name; @Required private String teacherName; @Required private String claName; private float score; public String getName() { return name; } public void setName(String name) { this.name = name; } public String getClaName() { return claName; } public void setClaName(String ClaName) { this.claName = claName; } public String getTeacherName() { return teacherName; } public void setTeacherName(String teacherName) { this.teacherName = teacherName; } public String getStudentId() { return studentId; } public void setStudentId(String id) { this.studentId = id; } public float getScore() { return score; } public void setScore(float score) { this.score = score; }}for (int i=0 ; i<10000 ;i++) { realm.beginTransaction(); StudentRealm realmStudent = realm.createObject(StudentRealm.class,"studentId"+i); realmStudent.setName("name"+i); realmStudent.setTeacherName("setTeacherName"+i); realmStudent.setClaName("setClaName"+i); realmStudent.setScore(90); realm.commitTransaction();}GreenDao定义orm模型结构并做10000次插入
@Entity()public class Student { @Id private String studentId; @NotNull private String name; private String claName; private String teacherName; private float score; @Generated(hash = 1491230551) public Student(String studentId, @NotNull String name, String claName, String teacherName, float score) { this.studentId = studentId; this.name = name; this.claName = claName; this.teacherName = teacherName; this.score = score; } @Generated(hash = 1556870573) public Student() { } public String getStudentId() { return studentId; } public void setStudentId(String studentId) { this.studentId = studentId; } @NotNull public String getName() { return name; } /** Not-null value; ensure this value is available before it is saved to the database. */ public void setName(@NotNull String name) { this.name = name; } public String getClaName() { return claName; } public void setClaName(String claName) { this.claName = claName; } public String getTeacherName() { return teacherName; } public void setTeacherName(String teacherName) { this.teacherName = teacherName; } public float getScore() { return score; } public void setScore(float score) { this.score = score; }}DaoSession daoSession = ((App) getApplication()).getDaoSession();StudentDao sd = daoSession.getStudentDao();for (int i = 0; i < 10000; i++) { Student s = new Student(); s.setStudentId("StudentId"+i); s.setClaName("getClaName"+i); s.setScore(90); s.setName("name"+i); s.setTeacherName("tn"+i); sd.insertOrReplace(s);}批量插入
Luakit没有提供批量插入接口。
active android批量插入10000条数据。
ActiveAndroid.beginTransaction();for (int i=0 ; i<10000 ;i++) { Student s = new Student(); s.studentId = "studentId"+i; s.name = "name"+i; s.teacherName = "teacherName"+i; s.claName = "claName"+i; s.score = 90; s.save();}ActiveAndroid.setTransactionSuccessful();ActiveAndroid.endTransaction();realm android批量插入10000条数据。
Realm realm = Realm.getDefaultInstance();realm.beginTransaction();for (int i=0 ; i<10000 ;i++) { StudentRealm realmStudent = realm.createObject(StudentRealm.class,"studentId"+i); realmStudent.setName("name"+i); realmStudent.setTeacherName("setTeacherName"+i); realmStudent.setClaName("setClaName"+i); realmStudent.setScore(90);}realm.commitTransaction();GreenDao批量插入10000条数据
DaoSession daoSession = ((App) getApplication()).getDaoSession();StudentDao sd = daoSession.getStudentDao();ArrayList<Student> ss = new ArrayList<Student>();for (int i = 0; i < 10000; i++) { Student s = new Student(); s.setStudentId("StudentId"+i); s.setClaName("getClaName"+i); s.setScore(90); s.setName("name"+i); s.setTeacherName("tn"+i); ss.add(s);}sd.insertOrReplaceInTx(ss);数据查询
Luakit做10000次查询，下面的代码是ios和android通用的。
local studentTable = Table("Student")for i=1,10000 do local result = studentTable.get:where({"studentId"..i},"studentId = ?"):all()endactive android做10000次查询。
for (int i=0 ; i<10000 ;i++) { List<Student> student = new Select() .from(Student.class) .where("studentId = ?", "studentId"+i) .execute();}realm android 做10000次查询。
for (int i=0 ; i<10000 ;i++) { RealmResults<StudentRealm> students = realm.where(StudentRealm.class).equalTo("studentId", "studentId"+i).findAll(); List<StudentRealm> list = realm.copyFromRealm(students);}GreenDao 做10000次查询
DaoSession daoSession = ((App) getApplication()).getDaoSession();StudentDao sd = daoSession.getStudentDao();for (int i = 0; i < 10000; i++) { List<Student> s = sd.queryBuilder() .where(StudentDao.Properties.StudentId.eq("StudentId"+i)) .list();}循环更新
Luakit做10000次更新。
local studentTable = Table("Student")for i=1,10000 do local result = studentTable.get:where({"studentId"..i},"studentId = ?"):update({name = "name2”})endactive android做10000次更新。
for (int i=0 ; i<10000 ;i++) { ActiveAndroid.beginTransaction(); Update update = new Update(Student.class); update.set("name = ?","name2") .where("studentId = ?", "studentId"+i) .execute(); ActiveAndroid.setTransactionSuccessful(); ActiveAndroid.endTransaction();}realm android做10000次更新。
for (int i=0 ; i<10000 ;i++) { realm.beginTransaction(); StudentRealm student = realm.where(StudentRealm.class).equalTo("studentId", "studentId"+i).findFirst(); student.setClaName("ClaName"+(i+1)); realm.copyToRealmOrUpdate(student); realm.commitTransaction();}GreenDao做10000次更新。
for (int i = 0; i < 10000; i++) { List<Student> s = sd.queryBuilder() .where(StudentDao.Properties.StudentId.eq("StudentId"+i)) .list(); s.get(0).setName("name2"); sd.update(s.get(0));}批量更新
Luakit没有批量更新接口。
active android批量更新10000条数据。
ActiveAndroid.beginTransaction();for (int i=0 ; i<10000 ;i++) { Update update = new Update(Student.class); update.set("name = ?","name2") .where("studentId = ?", "studentId"+i) .execute();}ActiveAndroid.setTransactionSuccessful();ActiveAndroid.endTransaction();realm android批量更新10000条数据。
realm.beginTransaction();for (int i=0 ; i<10000 ;i++) { StudentRealm student = realm.where(StudentRealm.class).equalTo("studentId", "studentId"+i).findFirst(); student.setClaName("ClaName"+(i+1)); realm.copyToRealmOrUpdate(student);}realm.commitTransaction();GreenDao批量更新10000条数据
ArrayList<Student> ss = new ArrayList<Student>();for (int i = 0; i < 10000; i++) { List<Student> s = sd.queryBuilder() .where(StudentDao.Properties.StudentId.eq("StudentId"+i)) .list(); s.get(0).setName("name2"); ss.add(s.get(0));}sd.updateInTx(ss);循环删除
Luakit做10000次删除操作。
local studentTable = Table("Student")for i=1,10000 do studentTable.get:where({"studentId"..i},"studentId = ?"):delete()endactive android做10000次删除操作。
for (int i=0 ; i<10000 ;i++) { ActiveAndroid.beginTransaction(); new Delete().from(Student.class).where("studentId = ?", "studentId"+i).execute(); ActiveAndroid.setTransactionSuccessful(); ActiveAndroid.endTransaction();}realm android做10000次删除操作。
for (int i=0 ; i<10000 ;i++) { realm.beginTransaction(); StudentRealm student = realm.where(StudentRealm.class).equalTo("studentId", "studentId"+i).findFirst(); student.deleteFromRealm(); realm.commitTransaction();}GreenDao做10000次删除操作。
for (int i = 0; i < 10000; i++) { List<Student> s = sd.queryBuilder() .where(StudentDao.Properties.StudentId.eq("StudentId"+i)) .list(); s.get(0).setName("name2"); sd.delete(s.get(0));}批量删除
Luakit没有批量删除接口。
active android批量删除10000条数据。
ActiveAndroid.beginTransaction();for (int i=0 ; i<10000 ;i++) { new Delete().from(Student.class).where("studentId = ?", "studentId"+i).execute();}ActiveAndroid.setTransactionSuccessful();ActiveAndroid.endTransaction();realm android批量删除10000条数据。
realm.beginTransaction();for (int i=0 ; i<10000 ;i++) { StudentRealm student = realm.where(StudentRealm.class).equalTo("studentId", "studentId"+i).findFirst(); student.deleteFromRealm();}realm.commitTransaction();GreenDao批量删除10000条数据。
ArrayList<Student> ss = new ArrayList<Student>();for (int i = 0; i < 10000; i++) { List<Student> s = sd.queryBuilder() .where(StudentDao.Properties.StudentId.eq("StudentId"+i)) .list(); ss.add(s.get(0));}sd.deleteInTx(ss);android 测试结果及分析
下面给出测试结果，表格中所有数据的单位是秒，即做10000次操作需要的秒数。
- 可以看到，active android各项性能都一般。
- 在使用批量接口的情况下GreenDao和Realm的性能比较好。
- 在使用批量接口的情况下Realm的性能尤其好，批量插入、查询、批量更改、批量删除都是Realm的性能最好，但是Realm的非批量接口性能较差，所有可以这样总结，如果代码高内聚，可以把数据操作代码入口都统一使用，Realm性能是最好的，但这对代码质量、模块设计有要求，当操作数据的代码到处都有，不能使用批量接口时，Realm的性能是不好的。
- Luakit没有提供批量接口，但从图中可以看出，Luakit的各项性能指标都是比较好的，而且对代码没有要求，即使操作数据的代码不内聚，也不会对性能有影响。
ios测试代码
Luakit是跨平台的，代码跟android一样，下面就不列了，只给出Coredata和 Realm ios
循环插入
Coredata 定义orm模型结构并做10000次插入
@interface Student (CoreDataProperties)+ (NSFetchRequest<Student *> *)fetchRequest;@property (nullable, nonatomic, copy) NSString *claName;@property (nullable, nonatomic, copy) NSString *name;@property (nonatomic) float score;@property (nullable, nonatomic, copy) NSString *studentId;@property (nullable, nonatomic, copy) NSString *teacherName;@endself.context = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSMainQueueConcurrencyType];for (int i=0; i<10000; i++) { Student *s = [NSEntityDescription insertNewObjectForEntityForName:@"Student" inManagedObjectContext:self.context]; s.studentId = [NSString stringWithFormat:@"studentId%d",i]; s.name = [NSString stringWithFormat:@"name%d",i]; s.teacherName = [NSString stringWithFormat:@"teacherName%d",i]; s.claName = [NSString stringWithFormat:@"claName%d",i]; s.score = 90; NSError *error = nil; [self.context save:&error];}Realm ios定义orm模型结构并做10000次插入
@interface StudentRLM : RLMObject@property NSString *studentId;@property NSString *name;@property NSString *teacherName;@property NSString *claName;@property float score;@endfor (int i=0; i<10000; i++) { [realm beginWriteTransaction]; StudentRLM *s = [[StudentRLM alloc] init]; s.studentId = [NSString stringWithFormat:@"studentId%d",i];; s.name = [NSString stringWithFormat:@"name%d",i]; s.teacherName = [NSString stringWithFormat:@"teacherName%d",i]; s.claName = [NSString stringWithFormat:@"claName%d",i]; s.score = 90; [realm addOrUpdateObject:s]; [realm commitWriteTransaction]; [realm beginWriteTransaction];}批量插入
Coredata 批量插入10000条数据。
for (int i=0; i<10000; i++) { Student *s = [NSEntityDescription insertNewObjectForEntityForName:@"Student" inManagedObjectContext:self.context]; s.studentId = [NSString stringWithFormat:@"studentId%d",i]; s.name = [NSString stringWithFormat:@"name%d",i]; s.teacherName = [NSString stringWithFormat:@"teacherName%d",i]; s.claName = [NSString stringWithFormat:@"claName%d",i]; s.score = 90; }NSError *error = nil;[self.context save:&error];Realm ios批量插入10000条数据。
[realm beginWriteTransaction];for (int i=0; i<10000; i++) { StudentRLM *s = [[StudentRLM alloc] init]; s.studentId = [NSString stringWithFormat:@"studentId%d",i];; s.name = [NSString stringWithFormat:@"name%d",i]; s.teacherName = [NSString stringWithFormat:@"teacherName%d",i]; s.claName = [NSString stringWithFormat:@"claName%d",i]; s.score = 90; [realm addOrUpdateObject:s];}[realm commitWriteTransaction];[realm beginWriteTransaction];查询
Coredata 做10000次查询。
for (int i=0; i<10000; i++) { NSFetchRequest *request = [NSFetchRequest fetchRequestWithEntityName:@"Student"]; request.predicate = [NSPredicate predicateWithFormat:@"studentId = 'studentId%d'"]; NSArray *objs = [self.context executeFetchRequest:request error:&error];}Realm ios做10000次查询。
for (int i=0; i<10000; i++) { RLMResults *results = [StudentRLM objectsWhere: [NSString stringWithFormat:@"studentId = 'studentId%d'",i]]; StudentRLM *s = results.firstObject;}循环更新
Coredata 做10000次更新。
for (int i=0; i<10000; i++) { NSBatchUpdateRequest *batchUpdateRequest = [[NSBatchUpdateRequest alloc] initWithEntityName:@"Student"]; batchUpdateRequest.predicate = [NSPredicate predicateWithFormat:@"studentId = 'studentId%d'"]; batchUpdateRequest.propertiesToUpdate = @{@"name" : @"name2"}; batchUpdateRequest.resultType = NSUpdatedObjectsCountResultType; NSBatchUpdateResult *batchResult = [self.context executeRequest:batchUpdateRequest error:&error]; NSError *error = nil; [self.context save:&error];}Realm ios做10000次更新。
for (int i=0; i<10000; i++) { [realm beginWriteTransaction]; RLMResults *results = [StudentRLM objectsWhere: [NSString stringWithFormat:@"studentId = 'studentId%d'",i]]; NSLog(@"results %lu",(unsigned long)[results count]); StudentRLM *s = results.firstObject; [s setName:@"name"]; [realm addOrUpdateObject:s]; [realm commitWriteTransaction];}批量更新
Coredata 批量更新10000条数据。
for (int i=0; i<10000; i++) { NSBatchUpdateRequest *batchUpdateRequest = [[NSBatchUpdateRequest alloc] initWithEntityName:@"Student"]; batchUpdateRequest.predicate = [NSPredicate predicateWithFormat:@"studentId = 'studentId%d'"]; batchUpdateRequest.propertiesToUpdate = @{@"name" : @"name2"}; batchUpdateRequest.resultType = NSUpdatedObjectsCountResultType; NSBatchUpdateResult *batchResult = [self.context executeRequest:batchUpdateRequest error:&error];}NSError *error = nil;[self.context save:&error];Realm ios批量更新10000条数据。
[realm beginWriteTransaction];for (int i=0; i<10000; i++) { RLMResults *results = [StudentRLM objectsWhere: [NSString stringWithFormat:@"studentId = 'studentId%d'",i]]; NSLog(@"results %lu",(unsigned long)[results count]); StudentRLM *s = results.firstObject; [s setName:@"name”]; [realm addOrUpdateObject:s];}[realm commitWriteTransaction];循环删除
Coredata 做10000次删除操作。
for (int i=0; i<10000; i++) { NSFetchRequest *request = [NSFetchRequest fetchRequestWithEntityName:@"Student"]; request.predicate = [NSPredicate predicateWithFormat:@"studentId = 'studentId%d'"]; NSBatchDeleteRequest *batchRequest = [[NSBatchDeleteRequest alloc] initWithFetchRequest:request]; batchRequest.resultType = NSUpdatedObjectsCountResultType; NSBatchUpdateResult *batchResult = [self.context executeRequest:batchRequest error:&error]; NSError *error = nil; [self.context save:&error];}Realm ios做10000次删除操作。
for (int i=0; i<10000; i++) { [realm beginWriteTransaction]; RLMResults *results = [StudentRLM objectsWhere: [NSString stringWithFormat:@"studentId = 'studentId%d'",i]]; StudentRLM *s = results.firstObject; [s setName:@"name"]; [realm deleteObject:s]; [realm commitWriteTransaction];}批量删除
Coredata 批量删除10000条数据。
for (int i=0; i<10000; i++) { NSFetchRequest *request = [NSFetchRequest fetchRequestWithEntityName:@"Student"]; request.predicate = [NSPredicate predicateWithFormat:@"studentId = 'studentId%d'"]; NSBatchDeleteRequest *batchRequest = [[NSBatchDeleteRequest alloc] initWithFetchRequest:request]; batchRequest.resultType = NSUpdatedObjectsCountResultType; NSBatchUpdateResult *batchResult = [self.context executeRequest:batchRequest error:&error];}NSError *error = nil;[self.context save:&error];Realm ios批量删除10000条数据。
[realm beginWriteTransaction];for (int i=0; i<10000; i++) { RLMResults *results = [StudentRLM objectsWhere: [NSString stringWithFormat:@"studentId = 'studentId%d'",i]]; StudentRLM *s = results.firstObject; [s setName:@"name"]; [realm deleteObject:s];}[realm commitWriteTransaction];ios 测试结果及分析
下面给出测试结果，表格中所有数据的单位是秒，即做10000次操作需要的秒数。
- 可以看到，Coredata除了批量插入性能是最好的以外，其他项性能都一般。
- Realm ios和Realm android性能非常相似，批量操作性能优异，但是非批量操作性能一般。可以这样总结，如果代码高内聚，可以把数据操作代码入口都统一使用，Realm性能是最好的，但这对代码质量、模块设计有要求，当操作数据的代码到处都有，不能使用批量接口时，Realm的性能是不好的。
- Luakit没有提供批量接口，但从图中可以看出，Luakit的各项性能指标都是比较好的，而且对代码没有要求，即使操作数据的代码不内聚，也不会对性能有影响。