PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteSome links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
AI-generated SQL should be treated as untrusted code, not as a ready-to-run answer. The safe workflow is to freeze the intended database and dialect, parse the statement, inspect its structure, ask the target database to perform a prepare or dry-run check, review an execution plan without running the query, test it under restricted permissions, and verify the returned data against business rules. Only the exact statement that passes those gates should reach production. Parsing catches malformed syntax, but it cannot prove that tables, columns, permissions, types, or business logic are valid in your database.
What does it mean to validate AI-generated SQL?
Validation is a sequence of independent checks that answer different questions. A parser asks whether the text can be understood in a particular SQL dialect. The database asks whether the referenced objects, types, permissions, and engine-specific features are valid. An execution-plan inspection reveals likely scans, joins, and costs. A controlled run shows what actually happens with real data, while assertions and human review determine whether the result answers the original question.
As an Amazon Associate I earn from qualifying purchases.
No single check is a safety guarantee. A statement can parse successfully and still reference a nonexistent column, return duplicate customers because of a many-to-many join, update rows when the request was read-only, or expose data that the requester is not allowed to see. Treat each gate as evidence of one property, not proof of every property.
Why can AI SQL parse correctly and still fail?
SQL grammar is only one layer of correctness. A parser generally does not know your current schema, grants, data distribution, row-level security policies, or the business definition of a metric. It may also accept constructs that a target engine does not implement or interpret in the same way.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
- Schema errors: a table or column was renamed, exists in another schema, or is absent from the snapshot the model saw.
- Dialect errors: syntax valid in PostgreSQL, MySQL, SQL Server, Oracle, or SQLite is not automatically portable. Functions, date arithmetic, identifier quoting, pagination, and boolean behavior differ.
- Semantic errors: the query runs but uses the wrong grain, date boundary, join key, timezone, or aggregation.
- Operational errors: an unbounded scan, sort, or Cartesian join can exhaust resources even when the result is logically correct.
- Security errors: a generated batch may contain a write, stacked statement, file-access function, or external call that was never requested.
SQLGlot’s project documentation makes the boundary explicit: “SQLGlot is a transpiler, not a validator.” Oracle’s SQL processing documentation similarly notes, “Some errors cannot be caught by parsing.” Those statements are practical rules: use a parser first, then obtain an authoritative check from the target database.
What should I record before asking an AI model for SQL?
Freeze the request before generation. A short specification prevents the model from quietly choosing a dialect, broadening the scope, or adding write operations.
- Engine and version: record the exact database product, major version, and compatibility mode.
- Dialect: state the source dialect and, if translation is required, the explicit target dialect. SQLGlot documentation recommends passing both when translating.
- Schema snapshot: include the database, schema, tables, columns, data types, keys, views, and relevant constraints as they exist at a known time.
- Output contract: specify columns, sort order, expected grain, acceptable nulls, and whether duplicate keys are allowed.
- Statement policy: decide whether only
SELECTis allowed or whether a controlled migration may include particular write statements. - Read and resource limits: state whether the request must be read-only, the maximum runtime, row count, and result size.
- Data handling: identify sensitive columns that must not be selected or returned to the model.
Keep this specification with the generated SQL. If the model later receives an error, provide the exact error and the relevant schema context rather than asking it to rewrite the query blindly.
How do I parse AI-generated SQL safely?
Parse every statement away from production, using the intended dialect. Capture structured errors, including line and column information, and reject text that cannot be parsed before any database connection is opened.
- Normalize the input for inspection. Preserve the original text, but identify comments, string literals, semicolons, and batch boundaries. Do not remove comments by a simple regular expression if doing so could alter a string literal.
- Select the dialect explicitly. Do not rely on a parser’s default dialect. If the SQL is being translated, identify both the source and target engines.
- Parse all statements. A parser that accepts the first statement does not make later statements safe. Treat a multi-statement batch as a separate policy decision.
- Store parse diagnostics. Save the exception type, message, line, column, and the original SQL so a reviewer can reproduce the failure.
- Stop on parse failure. Do not “try it anyway” against a live connection. Ask for a minimal correction after identifying the exact token and dialect assumption that failed.
For a local Python check, a dialect-aware library such as SQLGlot can provide an AST for inspection. An illustrative pattern is:
import sqlglot
from sqlglot import parse_one
sql = "SELECT customer_id, SUM(amount) AS total FROM sales GROUP BY customer_id"
try:
statements = sqlglot.parse(sql, read="postgres")
tree = parse_one(sql, read="postgres")
except Exception as error:
print(f"Parse failed: {error}")
else:
print(tree)
Use the actual dialect and library version deployed by your team. Parser success means the text has a recognizable syntactic shape; it does not establish that the target database will execute it.
What should I inspect in the parsed SQL tree?
AST inspection turns a generated string into an allowlist decision. Traverse the tree and compare what it contains with the request and schema snapshot.
Statement type and batch shape
For a read-only task, reject any unexpected INSERT, UPDATE, DELETE, MERGE, DROP, ALTER, CREATE, or COPY. Reject multiple statements unless batching is explicitly required and separately approved. Look for transaction-control commands, procedural blocks, and vendor-specific commands that can have effects outside a normal query.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
Referenced objects and columns
List every table, view, common table expression, column, alias, and schema qualifier. Resolve aliases to their source tables and compare names and data types with the frozen schema. A model may use a plausible column name that exists in a different version of the application.
Joins and predicates
Record each join type and its condition. Flag joins without an equality or range predicate, joins that connect unrelated keys, and outer joins whose filters in the WHERE clause accidentally turn them into inner joins. Confirm that tenant, account, status, and date restrictions appear at the correct stage.
Recommended Free Tools
Functions, casts, and side effects
Identify date functions, casts, user-defined functions, dynamic SQL, file-access functions, network or external-table calls, and functions documented as having side effects. Verify that every function exists in the target engine and is permitted for the execution role.
Projection, grouping, and limits
Check selected columns against the output contract. Ensure nonaggregated columns are grouped correctly, window functions use the intended partition and order, and a LIMIT or equivalent is present when a preview is requested. A limit controls returned rows; it does not necessarily prevent an expensive scan or sort.
How do I enforce a read-only allowlist?
Make statement policy an explicit gate rather than relying on a prompt saying “read-only.” Parse the statement and classify its top-level operation before sending it to a database connection.
| Policy | Allow | Reject or require separate approval |
|---|---|---|
| Read-only report | SELECT, approved common table expressions, approved read-only functions |
INSERT, UPDATE, DELETE, MERGE, DDL, COPY, procedural blocks, stacked statements |
| Migration | Only the named operation and target objects | Unrelated tables, dynamic SQL, privilege changes, destructive drops, unreviewed data movement |
| Diagnostic query | Read statements and the engine’s plan command | Commands that execute functions with side effects or write plan output to external locations |
Enforce the same policy at the database layer with a least-privilege role. Application-level parsing is useful for early rejection, but it is not a substitute for credentials that cannot perform forbidden operations.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →How do I ask the database for a semantic or pre-execution check?
After parsing and AST inspection, use the target engine or driver to perform a prepare, statement-validation, or vendor dry-run check. This is where object names, column existence, data types, permissions, and engine-specific rules are evaluated without intentionally returning the full result.
- Connect with the restricted role. Use the same role and database context that the eventual job will use, but never grant production write access merely to test a read query.
- Prepare or validate the statement. Use the driver’s prepare API or the vendor’s documented validation command. Keep parameter values separate from the SQL text.
- Capture the complete database error. Store the SQLSTATE or vendor code, message, position, object name, and server version.
- Resolve one cause at a time. Distinguish an unknown column from a permission failure or an implicit-cast problem. Do not let the model change joins and filters while the actual issue is a missing grant.
- Repeat after every edit. A correction that fixes a column name can introduce a new type or grouping error.
Oracle’s documentation distinguishes syntax checks from semantic checks such as object and column existence, which is why parser-only validation is insufficient. Prepare behavior differs by engine and driver, so document the exact product and version in your procedure.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
How do I use EXPLAIN without accidentally running the query?
Use the database’s plan facility to inspect access paths, joins, scans, estimated cardinality, and cost before execution. Start with plain EXPLAIN or the vendor’s non-executing plan API; do not add ANALYZE until you have verified its semantics.
- Read the join order: look for an unexpected Cartesian join, a very large intermediate result, or a table scanned before a selective predicate is applied.
- Check access methods: an estimated full scan may be appropriate for a small table but hazardous for a large one. Confirm that expected indexes or partitions are eligible.
- Inspect estimates: a large gap between estimated and known row counts can produce a poor plan even when the SQL is valid.
- Review sorts and materialization: grouping, window functions, and ordering may require substantial memory or temporary storage.
- Check for hidden execution: some functions, virtual tables, foreign-data features, or driver behaviors may perform work while preparing or explaining.
Supabase documents that EXPLAIN ANALYZE executes the query and returns actual runtime data. SQLite describes EXPLAIN as an interactive analysis and troubleshooting output and warns that prepare-time effects are not all neutralized by EXPLAIN. Therefore, never treat a plan command as a universal safety boundary; verify your engine’s behavior and inspect the statement for side-effecting functions first.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsHow do I test generated SQL in a controlled environment?
Run only after parsing, structural inspection, and database validation. The safest test uses a sandbox or masked fixture that resembles production while limiting what the query can read and how many resources it can consume.
- Choose the data target. Prefer a disposable sandbox populated with representative, masked data. If a production read is unavoidable, use a dedicated read-only role and confirm the data classification.
- Set transaction behavior. For statements that can write, begin a transaction and roll it back where the engine guarantees that all relevant effects are transactional. A rollback is not protection against external calls, sequence changes, messages, or nontransactional objects.
- Apply limits. Set a statement timeout, memory or workload cap where supported, maximum result bytes, and a row limit appropriate to the test.
- Use a cancellation path. Ensure the driver or job runner can cancel a runaway query and that an operator knows how to terminate the session.
- Capture evidence. Save the database version, role, schema revision, execution plan, timing, row count, and errors. Do not store sensitive result rows unnecessarily.
- Compare with a known answer. Test a small fixture with hand-calculated outcomes for joins, filters, nulls, dates, and aggregates.
Do not assume that wrapping a statement in a transaction makes it harmless. Permissions, timeouts, and a sandbox are the primary controls; rollback is an additional recovery measure for operations that are truly transactional.
How do I check whether the result is logically correct?
Semantic validation asks whether the query answers the user’s question, not merely whether it returns rows. Turn the expected behavior into assertions that can fail loudly.
Grain and uniqueness
State what one output row represents: one order, one customer per month, or one event. Check that the supposed key is unique when it should be, and investigate duplicate rows before accepting a join.
Row-count and range checks
Compare the result count with a plausible range based on the request and known data volume. Set bounds for numeric totals, percentages, and dates. A result of zero or an unexpectedly huge count can indicate a missing predicate or an over-restrictive one.
Null and completeness checks
Measure null rates for required fields and verify that a left join has not been unintentionally filtered. Check that date ranges include the intended endpoints and that timezone conversions do not move records across a reporting day.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
Join and aggregation checks
Test each join independently on a fixture containing unmatched rows and duplicate keys. Reconcile aggregates at a second grain, such as daily totals against the monthly total, and compare a sample of records with a trusted report.
Business invariants
Encode rules such as “a shipment cannot precede its order,” “a percentage is between zero and 100,” or “a tenant can see only its own rows.” These checks catch plausible-looking SQL that uses the wrong business definition.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
How do I secure AI-generated SQL against injection and unintended access?
Generated SQL is untrusted input even when a person asked the model to write it. Keep values parameterized, constrain identifiers through an allowlist, and validate at every trust boundary.
- Parameterize values: bind user-supplied dates, IDs, search text, and limits through the driver instead of concatenating them into SQL.
- Allowlist identifiers: map a user’s approved report field to a known column name; never accept arbitrary table or column text as a parameter substitute.
- Validate type, length, format, and range: Microsoft’s security guidance says, “Never build Transact-SQL statements directly from user input.” Its recommended validation dimensions apply to generated-query pipelines as well.
- Use least privilege: the execution role should have only the schemas, tables, columns, and operations required for the task.
- Inspect comments and stacking: comments can conceal instructions from a reviewer, while semicolons can introduce additional statements.
- Review file and external access: reject functions that read files, invoke network resources, call external procedures, or write to locations outside the approved result path.
- Protect secrets: do not send passwords, tokens, raw personal data, or unrestricted schema dumps to a model. Mask values and minimize schema context.
OWASP describes SQL injection as execution of a user-controlled query and notes that successful attacks can read, modify, or delete data and perform administrative operations. Parameterization and permissions reduce that risk; neither removes the need to inspect generated statements.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What should I do when the generated SQL fails?
Debug from the first failing gate and preserve the evidence. A repeatable record prevents the model from guessing and makes a human reviewable.
- Classify the failure. Is it a parse error, policy rejection, prepare error, plan problem, timeout, permission error, or semantic assertion failure?
- Save the original context. Keep the original prompt, generated SQL, target dialect, schema version, parser diagnostics, database error, plan output, and failed assertion.
- Reduce the query. Remove projections and joins until the smallest failing fragment remains. Validate that fragment independently.
- Ask for a minimal patch. Give the model the exact error, relevant schema definition, required output, and statement policy. Request a targeted change rather than a completely new query.
- Re-run every gate. A patched statement must be parsed, inspected, prepared, planned, tested, and asserted again. Never jump directly from a textual fix to production.
- Escalate ambiguous results. If the query’s business meaning is unclear, ask the request owner to define the grain, date semantics, or inclusion rules before editing SQL.
Common symptoms have different remedies:
| Symptom | Likely cause | Next check |
|---|---|---|
| Parser reports an unexpected token | Wrong dialect, malformed syntax, or unbalanced parentheses | Confirm engine and line/column location; parse with the explicit dialect |
| Unknown table or column | Stale schema context, wrong schema qualifier, or alias error | Compare every reference with the current schema snapshot |
| Permission denied | The role lacks access or a function requires elevated privileges | Review grants and remove unnecessary objects or functions |
| Query runs but returns too many rows | Missing predicate, duplicate-key join, or wrong grain | Inspect join cardinality and assert uniqueness at the intended key |
| Query times out | Large scan, bad join order, sort, or stale estimates | Read the non-executing plan and test narrower predicates in the sandbox |
| Numbers look plausible but disagree with a report | Different date boundary, timezone, null treatment, or metric definition | Reconcile on a small fixture and document the business rule |
Which validation tools should I choose?
Choose tools by the gate they can prove. A broad parser is useful for local inspection, while the target database remains authoritative for execution validity.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →| Tool capability | What it proves | Important limitation |
|---|---|---|
| Dialect-aware parser | Tokenization, syntax shape, parse errors, and an AST | Cannot prove schema, permissions, runtime behavior, or business correctness; intentional parser leniency may exist |
| AST walker or linter | Statement allowlists, object inventories, joins, predicates, functions, and projections | Needs a current schema and policy rules; it can miss engine-specific runtime behavior |
| Driver prepare or vendor validation | Engine-specific syntax, object and column existence, types, and permissions | Semantics and side effects vary by database and driver; it is not a business-logic test |
EXPLAIN or plan API |
Expected access path, joins, scans, estimates, and performance hazards | May not be side-effect-free in every engine; ANALYZE can execute the query |
| Sandbox execution | Actual runtime behavior under controlled data, credentials, and limits | Fixture data may not represent every production distribution or policy |
| Assertions and human review | Whether results satisfy the requested grain, ranges, invariants, and meaning | Requires an explicit expected outcome and an accountable reviewer |
When selecting a service or library, compare dialect coverage, strictness and error reporting, AST access, schema awareness, write blocking, prepare and dry-run support, plan integration, parameter handling, resource limits, auditability, and whether SQL leaves your environment. A tool that sends statements to a remote service may be unsuitable for sensitive schemas even if its parser is capable.
What does a complete pre-production checklist look like?
Use this checklist for each generated statement and retain the result with the change or report definition.
- Target engine, version, source dialect, and target dialect are recorded.
- Schema snapshot and schema revision are identified.
- Requested output, row grain, filters, date and timezone rules, and allowed statement types are explicit.
- Every statement parses with the intended dialect, and diagnostics are saved.
- AST inspection found only approved statement types, schemas, objects, columns, functions, joins, and predicates.
- User values are bound parameters; dynamic identifiers come from an allowlist.
- The restricted role can prepare or validate the statement without granting unnecessary access.
- A non-executing plan was reviewed; no unexpected Cartesian join, scan, sort, or external operation remains.
- The statement ran in a sandbox or under documented read-only limits, timeout, and result cap.
- Row counts, key uniqueness, null rates, date boundaries, aggregates, and business invariants passed.
- The exact reviewed SQL is versioned or hashed, with approval and test evidence retained.
When is it safe to run the final statement?
Run it only after the exact text has passed all applicable gates and the approval covers that exact version. Promote the reviewed statement rather than regenerating it at deployment time. Record its hash or version, the dialect and schema revision, the execution role, the plan evidence, the assertions, and the approver.
If any material input changes—database version, schema, permissions, model-generated text, parameters, or business definition—treat the statement as new and repeat the checks. This discipline is what turns AI assistance into a controlled development aid instead of an unreviewed production code path.
Free tools Windows power users keep installed
One-click scans. No signup required.
Frequently Asked Questions
Can I validate SQL without connecting to the database?
You can validate syntax, statement shape, and many policy rules locally with a dialect-aware parser and AST inspection. You cannot establish that objects, permissions, types, or engine-specific semantics are valid without an authoritative check from the target database or an equivalent isolated instance.
Should I let an AI model see my full database schema?
Share the smallest schema description needed to write the query, and remove secrets and sensitive row data. Table and column names can themselves be confidential, so use masked or synthetic names when the model does not need the originals.
Does wrapping a generated UPDATE in a rollback make it safe?
Only effects covered by the database’s transaction guarantees can be undone reliably. External calls, nontransactional objects, sequence changes, messages, and resource exhaustion may persist, so use a sandbox and a role that cannot perform unapproved writes.
What is the safest way to debug a query that times out?
Keep the original statement, apply a timeout and cancellation path, and inspect a non-executing plan first. Reduce the query to the smallest slow fragment, test narrower predicates in a representative sandbox, and then rerun semantic checks after any optimization.
How should I store evidence that an AI query was reviewed?
Keep the prompt, exact SQL, dialect, schema revision, parser and database diagnostics, plan output, test parameters, assertions, approval, and a hash or version of the statement. Restrict access to result samples and redact sensitive values from logs.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
