NOCLAU GUIDE
Copy and Paste Text Troubleshooting
Use a three-sample comparison to investigate text mismatches, isolate changes, and validate a repair.
When copied text looks correct but behaves differently, the fastest useful investigation is usually a comparison of a few controlled samples. Repeatedly running a cleaner without a hypothesis can erase evidence while leaving the cause untouched. This guide provides a practical worksheet for finding which step changes a value, testing a proposed repair, and writing a bug report another person can reproduce. It applies to identifiers, short labels, and prose, with extra care whenever structured data or several writing systems are involved.
Scroll the table sideways to see every column.
| Sample | How to create it | What to record |
|---|---|---|
| A: typed | Type expected value directly | Does the operation succeed? |
| B: source | Paste from original source | Which behavior differs from A? |
| C: changed copy | Apply one explained edit | Does it match A after saving? |
Define expected behavior precisely
Write down the exact operation that fails. A product code might fail a lookup, a phrase might not appear in search results, or a label might wrap at an unexpected place. Include the application and the expected outcome. Saying that text is weird is understandable, but not enough to distinguish a content problem from a font, style, import, or validation problem.
Keep an untouched copy of the source and a separate working sample. If the source is long, choose a small portion that still reproduces the behavior. Preserve the characters immediately around the suspected location. An isolated mark may not reproduce a joining or direction issue because its effect depends on the surrounding sequence. Use invented content when the real sample contains private information.
Create three versions of the same value
Version A is the expected value typed directly in the destination. Version B is pasted from the original source. Version C is the inspected or cleaned working copy. Perform the same operation on all three. If A succeeds and B fails, there is a meaningful difference somewhere in the copying path. If C succeeds after one explicit edit, that edit is a candidate explanation.
This experiment is more informative than cleaning a large document and observing that it seems better. It also catches a different possibility: all three versions fail because the destination expects another format. In that case, changing hidden characters may be irrelevant. Review the documented field format, any length limit, and the actual comparison rules before making more edits.
Inspect selected characters and ordinary whitespace
Paste Version B into the analyzer and record its findings. A U+200B inside an identifier, a U+00A0 in a label, or a directional control can explain some mismatches. Turn on Show spaces and Show tabs to examine ordinary whitespace in the preview. These visual controls do not insert their dots or arrows into the copied output.
An empty issue list has limits. The scanner does not report ordinary leading or trailing spaces as special findings, does not list every Unicode character, and does not compare two strings. Check the beginning and end of the value manually. If a long report appears, remember that the badges show the first thirty findings; split the sample into smaller sections to inspect the relevant location.
Change one variable at a time
If the evidence points to an accidental zero-width separator, try conservative cleanup on a copy. If it points to a special space, test Normalize spaces separately. Keep dashes unchanged unless punctuation is the specific issue. Multiple simultaneous edits make it harder to identify which one solved the problem and which ones introduced unrelated changes.
Use a short record such as Original: AB[U+200B]12; Edit: remove U+200B; Result: AB12; Test: exact lookup succeeds. The bracketed notation is for explanation, not literal input. This record captures a causal hypothesis you can repeat. If the test fails, restore the original and revise the hypothesis rather than continuing to remove more categories of characters.
Investigate differences beyond hidden marks
Two visually similar strings can differ in case, punctuation, or the letters used. Accented text may also use different underlying representations. Unicode normalization is a separate operation described in the normalization specification; noclau does not perform general NFC or NFKC normalization. Do not assume the Normalize spaces checkbox performs either of those transformations.
If the problem involves an accent or a visually similar letter, use a tool that exposes the full sequence and compare it with the expected value. Compatibility normalization can also change distinctions you may want to preserve, so it should follow the requirements of the receiving system. For a person's name or published quotation, do not change the content simply to make an unexplained comparison succeed.
Check the rich-text and clipboard boundary
A plain-text tool cannot preserve all the structure in a document editor. If the problem involves a heading level, hyperlink destination, list indentation, or table layout, inspect the destination's paste settings. Compare a normal paste with the editor's plain-text paste option using a harmless sample. Determine which behavior you actually need before rebuilding formatting.
The copy button writes the result after you press it. If browser permissions prevent that action, the tool asks you to select and copy manually. Always confirm the copied content by pasting into the destination before replacing an important value. Device clipboard history and browser extensions are separate from the cleaner, so account for them when investigating unexpected copied content.
Validate the saved result, not just the preview
Some problems appear only when a field is saved, imported, or published. Run the final operation and then reopen the value. Compare what was stored with what you pasted. A destination may apply its own formatting or validation after input, which would explain why repeated cleanup seems to have only a temporary effect.
For a publishing task, inspect both the saved editor text and the rendered page. For a data import, test one record and verify its columns and values before running a full batch. The cleaner does not validate CSV structure, source code, or external link safety. A repaired text fragment still needs the normal acceptance checks for the system receiving it.
Share a useful minimal report
A good report includes the source and destination applications, a short invented example, exact steps, the selected rule, and the expected versus actual result. Include the code point and displayed position when relevant. Explain whether direct typing succeeds and whether the problem appears before or after saving. This gives the recipient a clear reproduction without asking them to guess your environment.
Avoid sending entire private documents or account values. Keep the original locally until the issue is resolved, and retain the smallest accepted transformation as a regression example for your workflow. The aim is to understand and fix one reproducible failure. A report with a clear test is more useful than a general claim that copied text contains a watermark.
References and further reading
Examples on this page are constructed demonstrations, not measurements of how often a platform inserts characters.