Manage six types of databases in one place on Windows·Ubuntu·macOS.
Community is freeProvides SQL editing and basic heterogeneous DDL·data transfer. Pro is paidAdds incremental transfer, scheduling, and Oracle PL/SQL migration.
Querydog is an SQL agent that allows different types of databases to be integrated and used within a single program. From SQL writing to schema management, data copying, backup·scheduling — It is designed to be easy and convenient for database beginners.
Installation files for Windows · Ubuntu Linux · macOS
Oracle migration verification · Measured comparison of DB transfer · User manual
Move table structures and data from one database to another. DDL transfer and data copying are executed separately, and some format differences are automatically converted. After moving, Verification Reportyou can check the transfer results and comparison range. Import/export of Excel·CSV and impact analysis are also provided.
You can place two databases side by side and check where and how they differ. We summarize whether a table exists only on one side, how the structure (DDL) differs even if the names are the same, and how much the record count differs. This is especially useful when aligning the structures of production and development environments.
Execute two SQL queries separately and compare the results row by row. We show which rows have changed, which exist only on one side, and which have disappeared, and you can export this to Excel. You can set it to ignore case, leading/trailing spaces, and the time part of dates, making it easy to filter out "essentially identical data."
Both editions use the same installation file. You can start with the free Community edition, and if you have a Pro license, you can use additional features in the same program.
COMMUNITY
Free
Functions needed for routine SQL tasks and basic data transfer.
PRO
Paid · License required
Adds repetitive tasks and advanced analytics features to all Community functions.
| Features | Community (Free) | Pro (Paid) |
|---|---|---|
| DB connection, SQL editing, table management | Available | Available |
| Basic DDL and data transfer (homogeneous and heterogeneous) | Available | Available |
| Transfer result reports and query result comparison | Available | Available |
| Incremental transfer and scheduled execution | Pro license required | Available |
| DB structure comparison and Oracle migration | Pro license required | Available |
| Cross DB queries, SQL debugger, visual query builder | Pro license required | Available |
Community refers to the free usage scope and does not mean open-source license. Please separately check the cost of external AI services and detailed support scope by DB type. Basis for transfer functions and detailed comparison · User manual
You can work the same way regardless of which database you use.
Some data types are automatically converted during transfer between DBs. The support scope may vary depending on DB type and version, so check the generated DDL and conversion results before transfer.
Connection passwords are not left in the connection setting file(Connection.ini) and are separated into a separate file in the user account folder for storage.Even if you share the setting file with colleagues or upload it to version control, the password does not go along with it.
We disclose transfer methods, type conversions, free range, and validation limits.
Reads multiple rows from the source DB and loads them into the target DB in batches. After checking the DDL preview, execute structure transfer and data copy separately. Transfer screen and mode view
Oracle NUMBER(p,s)There are automatic rules, such as converting to MariaDB's DECIMAL(p,s)types. Before executing DDL, you can check the source and target types for each column and warnings about downscaling, and modify and save the target types. Actual value lossless conversion must be verified after transfer. View type review procedures
Community provides free basic DDL and data transfer for homogeneous and heterogeneous databases. Incremental transfer, scheduled execution, permission copying, and Oracle PL/SQL migration are in the Pro range. View features by edition
If source reading is interrupted, the process stops at the current table, and subsequent tables are marked as unprocessed. Tables with failed saves or truncated values are not considered successful. After restoring the connection, you can retry with key comparison synchronization, and matching PK/UNIQUE keys are required on both sides. View retry procedures
You can check transfer reports and comparison screens. Matching row counts alone do not guarantee that all values and FKs match. MariaDB Large-scale Transfer Actual Measurement ComparisonChecked the number of rows per table and preservation of existing indexes and DDL. This was not a test to verify the overall value consistency. View implementation basis and verification scope
In isolation tests of Querydog 1.0.45 source, 40/40 collected objects were compiled to MariaDB DDL. Among these, the specified input results of 10 newly created test routines and 259 input results of 4 existing business date functions were confirmed. DDL compilation 40/40 does not indicate overall business migration success rate or data transfer completion rate.
TYPE·PIPELINED functions require a change in the calling method, and the full business data, transactions, and external call sites have not yet been verified. View test conditions, improvement history, and remaining constraints
Provides basic features for autocomplete, color highlighting, and SQL formatting. Autocomplete identifies which table you are currently querying and shows only the columns that can be used in that context. Executed queries are automatically saved in history (Ctrl+Shift+H), and frequently used queries can be saved as bookmarks. CASE, MERGE The same sentence patterns can be immediately used as code snippets.
You can follow a stored procedure or PL/SQL line by line to see how the values change (except for SQLite). When tracing, it executes a temporary copyinstead of touching the original, so there is no worry about changing objects in operation. It also provides a visual query builder and the ability to join tables from different databases into a single SQL (read-only).
Columns containing photos, documents, or long texts are initially ...folded, and only expand when double-clicked. This prevents the screen from becoming heavy.
Even if modifying a row fails, only that row is rolled back, and rows corrected by others first are highlighted in orange to indicate a conflict.
you can preview before sending.
Ctrl+Shift+Pshows in a tree format the order in which your SQL is processed within the database when you press it.
Steps with high cost or reading many rows are highlighted in red or orange, so you can immediately see where adjustments are needed.
Actual ExecutionWhen turned on, you can check actual values, not just estimates (real measurement mode excluded for SQLite).
Before executing (F5), the program first scans the SQL. If the syntax is wrong, it catches it and notifies you before sending it to the database.
DELETE without WHERE·UPDATE without WHERE·TRUNCATE·DROPFor dangerous statements like this
It asks once more, 'Do you really want to execute?'
Displays the count, size, lock status, compression status, and description in one line in the table list.
With one right-click, you can copy and paste the table, and if the same name exists, (1), (2)It will automatically add a number as shown.
Within the same connection, you can also drag with the mouse to move to another schema.
In the table designer, you can edit columns, indexes, foreign keys, and CHECK constraints, and preview the SQL that will be generated.
It maintains the ascending/descending settings of foreign keys and indexes that reference other schemas, and will warn you before applying if the number of foreign key columns does not match.
Preserves the disabled CHECK state in MSSQL and the NOT VALID status in PostgreSQL, and generates SQL to change the PostgreSQL tablespace.
Unsupported table option changes are blocked before application. View how to use the table designer
The above structure modifications were verified in the source and automated tests of Querydog 1.0.58. The results of applying to actual production DBs need to be separately validated.
Automatically draws a diagram showing how tables are connected. You can drag and arrange boxes, and the layout you create can be saved and reused. The completed diagram can be exported as SVG, PDF, or PNG to be directly used in meeting materials, and conversely, you can import an SQL file (DDL) to create a diagram.
If you explain it verbally, it becomes SQL.
If you explain in everyday language, like "Show me the top 10 products in sales last month," AI will generate the SQL statement for you.
You can choose Google Gemini, OpenAI GPT, Anthropic Claude, or even a local AI (Ollama) running on your own computer.
Run AIWhen you run a query, it also tells you what part is causing an error and how to fix it.
If you ask for tuning, like "Check why this query is slow," the execution plan just extracted and the index information on that table are also provided to the AI. So you can receive informed responses, such as "It's scanning the entire table because there's no index."
We analyze which tables are heavy and which indexes are fragmented (fragmentation rate) and automatically generate the SQL needed for optimization (ANALYZE, VACUUM, REINDEX, etc.). We measure how much the character columns are actually used and suggest an appropriate size to reduce wasted space.
It automatically records who created which user, who granted or revoked permissions, and which SQL was executed and when. You can view it directly on the screen or export it to CSV for storage, so you can use it with confidence in work environments where change tracking and accountability are required.
If you register backup, query, and data transfer tasks on a schedule, they will automatically execute at the designated time. A mode that runs with commands only, without a screen, is also provided, so you can integrate it into your company's automation system.
You can choose between a dark or light screen (9 themes) and adjust the font and row height yourself. Windows (dockable) can be freely arranged and will appear in the same place even after restarting. The core parts of the program are distributed in compiled form, making them difficult to tamper with.
We compared the time it takes to copy large amounts of data between MariaDB, Oracle, PostgreSQL, SQL Server (MSSQL), and SQLite. You can check whether DBeaver’s bulk load option was used, the improvement results by Querydog version, and the verification scope together.
Record update: October 29, 2026 · Querydog source tests version 1.0.11~1.0.24 · DBeaver 26.2.3
This is the actual measurement record for 11 transfer directions conducted so far. Tests with different numbers of rows, indexes, and environments are not compared in a single speed ranking. The DBeaver column shows previous measurements for that direction, and unmeasured or failed cases were not converted into time. 37 tests involve 6,387,550 rows, and 36 tests involve 4,444,314 rows excluding one large table.
| Transfer direction | Querydog version | Querydog time | DBeaver record | Data and index conditions | Verification |
|---|---|---|---|---|---|
| MariaDB → MariaDB | 1.0.11 | 5 minutes 50.0 seconds | Approximately 5 minutes 22~28 seconds | 37 items · 6,387,550 rows · 45 indexes | Verify row count and preservation of existing DDL/indexes |
| MariaDB → Oracle | 1.0.12 | 4 minutes 21.05 seconds | Approximately 4 minutes 5~34 seconds | 37 items · 6,387,550 rows · 0 indexes | Row count matches after adjusting character length |
| Oracle → MariaDB | 1.0.15 / retention code | 2 minutes 41.25~41.26 seconds | 3 minutes 0.03~4.13 seconds | 37 items · 6,387,550 rows · 0 indexes | Measured twice · row count matches · no verification of all values |
| MariaDB → PostgreSQL | 1.0.20 | 2 minutes 54.70~55.88 seconds | 2 minutes 51.22 seconds | 36 items · 4,444,314 rows · 0 indexes | Measured twice · Number of rows and separate content hash match |
| PostgreSQL → MariaDB | 1.0.21 | 3 minutes 1.89~2.27 seconds | 4 minutes 49.79 seconds | 36 items · 4,444,314 rows · 42 indexes | Measured twice · Matches checksum with DBeaver loaded results |
| MariaDB → MSSQL | 1.0.22 | 9 minutes 42.26 seconds | Bulk load test failed | 36 items · 4,444,314 rows · 34 primary key indexes | Querydog total row count · Sample 4,421 rows match |
| MSSQL → MariaDB | 1.0.22 | 3 minutes 45.81 seconds | Not measured | 36 items · 4,444,314 rows · 34 primary key indexes | Querydog total row count · Sample 4,421 rows match |
| Oracle → PostgreSQL | 1.0.22 | 2 minutes 55.01 seconds | Previous bulk load 3 minutes 9.76 seconds | 36 items · 4,444,314 rows · 0 indexes | Recent test row count · Sample 4,421 rows match; No simultaneous re-measurement with DBeaver |
| PostgreSQL → Oracle | 1.0.23 | 3 minutes 27.17 seconds | Not measured | 36 items · 4,444,314 rows · 0 indexes | Row count · Sample 4,421 rows match |
| MariaDB → SQLite | 1.0.24 | 3 minutes 23.45 seconds | Not measured | 36 items · 4,444,314 rows · 0 indexes · Local | Row count · Sample 4,421 rows match |
| SQLite → MariaDB | 1.0.24 | 2 minutes 40.05 seconds | Not measured | 36 items · 4,444,314 rows · 0 indexes · Local | Row count · Round-trip sample 4,421 rows match |
Below are the major versions and retest history. The initial comparison and error history of MariaDB forward and Oracle forward have been preserved in the existing detailed record at the bottom of this page.
| Test | Version · Classification | Time Changes | Interpretation and Limitations |
|---|---|---|---|
| Oracle → MariaDB | 1.0.12 → 1.0.13 → 1.0.14 → 1.0.15 | 7 minutes 50.16 seconds → 4 minutes 59.58 seconds → 4 minutes 36.16 seconds → 2 minutes 41.26 seconds | 24 DDL errors in 1.0.12 measured after calibration for testing. 37 DDLs successfully created in 1.0.13. Not a full precision verification. |
| Oracle → MariaDB additional candidate | 1.0.16 Test | 2 minutes 38.44 seconds / 3 minutes 7.45 seconds / 2 minutes 54.89 seconds | The effect was not reproduced in repetitions, so candidate change was reverted. Retest of the adopted code was 2 minutes 41.25 seconds. The fastest candidate value is not used as the product representative value. |
| Oracle → PostgreSQL | 1.0.16 → 1.0.17 → 1.0.18 → 1.0.22 | 31 minutes 46.76 seconds → 6 minutes 23.58 seconds → 3 minutes 23.92 seconds → 2 minutes 55.01 seconds | History of applying COPY automatic loading, prefetching, etc. Each point is a test, not an experiment separating the individual effect of each change. |
| Oracle → PostgreSQL / DBeaver | 26.2.3 | Basic 9 minutes 14.24 seconds → Bulk loading 3 minutes 9.76 seconds | Bulk loading and multi-row INSERT 500 turned on together. CSV bulk import confirmed in log; individual effects of the two options are not separated. |
| MariaDB → PostgreSQL | 1.0.19 → 1.0.20 | 3 minutes 43.14 seconds → 2 minutes 54.70~55.88 seconds | Both measurements with the same final code are within 20 seconds of previous DBeaver actual measurements. |
| PostgreSQL → MariaDB | 1.0.20 → 1.0.21 | 5 minutes 16.82 seconds → Average 3 minutes 2.08 seconds | Applied server cursor and prefetching. In both cases, 42 indexes and checksum matched DBeaver results. |
| MariaDB → MSSQL | Previous test → 1.0.22 | Previous stop → 9 minutes 42.26 seconds | The previous 50 minutes or more was an estimate, not an actual measurement. General table OPENJSON input, data type and date handling improved. 8 auxiliary indexes are outside the scope of the test. |
| MariaDB → MSSQL / DBeaver | Microsoft JDBC 13.4.0.0 | No completion time | Used Bulk Copy, commit every 20,000 rows. Stopped due to source data read error; 555,077 rows loaded. The time until the interruption is not compared with the successful completion time. |
| PostgreSQL → Oracle | 1.0.22 → 1.0.23 | 5 minutes 12.84 seconds → 3 minutes 27.17 seconds | 105.68 seconds (33.8%) reduction observed. Server cursor, one-batch prefetch, general batch of 10,000 rows. |
| MariaDB → SQLite | 1.0.23 → 1.0.24 | 3 minutes 45.02 seconds → 3 minutes 23.45 seconds | 21.57 seconds (9.6%) reduction observed. Optimized general value conversion checks; streaming and prefetch not applied. |
| SQLite → MariaDB | 1.0.23 → 1.0.24 | 2 minutes 41.35 seconds → 2 minutes 40.05 seconds | Reverse regression check. Not a result of separate speed improvements. |
Download public summary data (JSON) · Existing detailed comparison of MariaDB · Existing detailed comparison of Oracle
37 tables · 6,387,550 rows · Copy test maintaining 45 existing indexes
Measurement Date: · Internal measurement by Windsoft · Each program run once from Windows GUI
Data was copied from the same source MariaDB to the same remote MariaDB. Before each run, only the data in the target tables was cleared while maintaining the existing table structure and indexes. In this run Querydog took 5 minutes 50.0 seconds, DBeaver about 5 minutes 22~28 seconds, so Querydog took about 22~28 seconds (7~9%) longer.
| Comparison Items | Querydog Community 1.0.11 | DBeaver 26.2.3 |
|---|---|---|
| Duration | 5 minutes 50.0 seconds (350.0 seconds) | About 5 minutes 22~28 seconds (322~328 seconds) |
| Time Measurement Method | Internal measurement from start to finish of transfer worker | Observed range from execution command to UI task completion |
| Tables / Total Rows | 37 / 6,387,550 rows | 37 / 6,387,550 rows |
| Row count check per table | All 37 match the source | All 37 match the source |
| Existing Indexes | 45 maintained | 45 maintained |
| Table Definition (DDL) | Matches the pre-run standard | Matches the pre-run standard |
| Error Check | Report: Success 37, Skipped 0, Error 0 | No error dialog observed Separate error count not secured |
Note: The separate execution record of the previous Querydog 1.0.10 was 6 minutes 18.2 seconds. This version 1.0.11 reduced that record by about 28.2 seconds (7.46%). The measured version and the currently available installation file version may differ.
Implementation basis and verification scope of inter-DB transfer · User manual for transfer
Measurement date: October 27, 2026 · 37 default tables · Original 6,387,550 rows · Target Oracle 21c XE
Querydog 1.0.12 completed the transfer of 37 tables in 4 minutes 21.05 secondsand in the post-check, the row count of all tables matched the source. This result falls within the previous DBeaver measurement range of about 4 minutes 5~34 seconds. Since the execution method and drivers are different, this is not a test to determine speed superiority between products.
| Items | DBeaver 26.2.3 | Querydog 1.0.11 Before correction | Querydog 1.0.12 After correction |
|---|---|---|---|
| Duration | Approximately 4 minutes 5~34 seconds (245.125~273.854 seconds) | 5 minutes 29.09 seconds (329.09 seconds) | 4 minutes 21.05 seconds (261.05 seconds) |
| Transfer results | Row count match for 37 tables | Success: 33 / Fail: 4 | Success: 37 / Errors: 0 / Warnings: 0 |
| Actual loaded row count | 6,387,550 rows | 5,997,256 rows | 6,387,550 rows |
| Verification of row count by table | All 37 match | 33 match / 4 not loaded | All 37 match |
| Target Index | 0 items | 0 items | 0 items |
| Execution/Measurement Method | GUI transmission / observation period of execution/termination screen | GUI transmission / internal worker measurement | Using an independently executed actual transmission worker like GUI / internal time measurement |
The first execution of DBeaver occurred due to differences based on the number of characters and bytes ORA-12899that occurred. After correcting the target string column length based on the character count, the result of retransmission was recorded in the table above. Querydog also used the same corrected table.
In Querydog 1.0.11, 4 tables failed due to case and quotation handling issues in Oracle column names. In 1.0.12, it was modified to preserve the actual column names, reduced repetitive value conversions and row copying, and increased the general table input batch from 5,000 rows to 20,000 rows.
The observed time after modification was approximately reduced by 68 seconds (20.7%). However, the executions before modification include failures for some tables, whereas the executions after modification were all successful, so it cannot be interpreted as a pure performance improvement rate for the same amount of successful executions.
Transfer comparison for MariaDB targets · Implementation basis and verification scope
You can continue tasks such as connecting to multiple databases, executing SQL, and checking table structures and data in Querydog. It is a database management tool for users on Windows, Ubuntu, and macOS handling Oracle, MySQL, MariaDB, PostgreSQL, SQL Server, and SQLite.
Check SQL editing and execution results, schema comparison, data copying and verification, backup, and scheduling functions according to actual tasks. Available features may vary depending on the type and version of the DB.
Review alternatives to Navicat and DBeaver, and guide on using Querydog for each DB. · Check DB migration functions and verification scope.
Check the current distribution status from the installation file list, review features and verification scope, and then select.
Check free download User manualInstallation files for Windows · Ubuntu Linux · macOS