# Jianlai tutorial 30: record host undo-redo states without assuming one character per action

Language: en
Canonical: [Jianlai tutorial 30: record host undo-redo states without assuming one character per action](<https://allluckly.cn/en/guides/jianlai-host-undo-redo-observed-text-state-chain/>)
中文：[中文 HTML](<https://allluckly.cn/guides/jianlai-host-undo-redo-observed-text-state-chain/>)
English: [English HTML](<https://allluckly.cn/en/guides/jianlai-host-undo-redo-observed-text-state-chain/>)
Published: 2026-11-17T00:00:00+08:00
Modified: 2026-11-17T15:44:18+08:00
Author: Bison

Use Jianlai to revise the minutes in a safe practice copy, then record text states through the host’s actual Undo and Redo facilities. Deliver before, after and each observed intermediate state. Edit grouping is determined by actual host behavior, not an assumed one-character rule or invented Jianlai-specific controls.

English translation of the [Chinese original](<https://allluckly.cn/guides/jianlai-host-undo-redo-observed-text-state-chain/>), translated on 2026-11-17. Original publication dates, attribution, code and version notes are retained.

Use Jianlai to revise the minutes in a safe practice copy, then record text states through the host’s actual Undo and Redo facilities. Deliver before, after and each observed intermediate state. Edit grouping is determined by actual host behavior, not an assumed one-character rule or invented Jianlai-specific controls.

## 1. Choose a host safe for rollback practice

Create a nonprivate ordinary text copy outside real outgoing messages, shared drafts and working business forms. First identify whether the current host or system has available Undo and Redo actions. Names, menus or gestures depend on device and host. If no reliable action is found, prepare only before-and-after text and mark history-chain testing unperformed. Apple’s editing guide describes system undo-redo, but does not guarantee identical facilities in every custom field or buttons on a Jianlai toolbar. Annotate host, document identity and actual controls instead of repeatedly tapping unknown locations. Retain a separate baseline so no rollback can affect the sole official original.

## 2. Enter and retain the baseline

Confirm Jianlai and enter the baseline sentence below, preserving the Chinese comma and full stop as shown while checking the time colon, identifier space and alphanumeric text literally. The meeting time is original exercise content, not a created appointment or reminder. Commit composition so the baseline is body text, then retain a comparable text copy or screenshot and record that it is committed. For the history experiment, stay in the same editing session where possible rather than first closing the document to prove persistence: reopening may change history availability in different hosts. Keep annotations in a separate explanation area so their edits do not enter the current undo history and obscure which action each step reverses.

```
会议 09:05，编号 B07。
```

## 3. Replace only the minutes and check the result

Use the host’s real selection or insertion point to change zero-five after the colon to five-zero, preserving everything else. Enter replacement characters through Jianlai, commit and compare with the revised sentence below. In particular, the hour must remain zero-nine, the identifier unchanged and the Chinese punctuation not overwritten by an oversized selection. If the operation requires deletion followed by typing, record those actual actions rather than simplifying them into an unobserved atomic replacement. The revised sentence is the edit endpoint, not evidence that Undo has been tested. Retain it separately as the textual reference for later Redo. If the edit is initially wrong, fix it and record additional actions instead of hiding entry errors behind subsequent history differences.

```
会议 09:50，编号 B07。
```

## 4. Record each actual Undo result

Invoke the previously identified host or system Undo once, then stop and record the complete visible body as the first undo state. Note any special insertion-point or composition state too. Do not prefill an expectation of one character removed or an immediate return to baseline. The host may group several inputs together, and deletion and insertion may be grouped differently. If the baseline is reached, stop undoing rather than affecting earlier content. If an intermediate state appears, record it before deciding whether one further step is safe. If an action also removes baseline entry, record that the group crossed the baseline and stop. Blank text is not a correct restoration, and unknown history is not a reason for repeated blind Undo.

## 5. Observe Redo only where available

From a recorded undo state, check whether Redo is actually available. If clearly identified, invoke it once, record the complete body and compare with the retained revised target. If not yet at that target, continue only where availability and scope remain clear, recording every state without predetermining a count. Avoid new typing or switching documents first, since a new edit or session change may alter redo history. If that already happened, record the condition change instead of automatically blaming Jianlai for unavailable Redo. Mark “this return reached the revised sentence” only after observing the target with neighbors intact. If unsupported or history has cleared, retain evidence and the source copy and mark the test incomplete rather than claiming unseen content was restored.

## 6. Separate composition from committed history

If Undo changes only a candidate or uncommitted spelling, record a composition-layer change rather than restored body history. Requiring committed baseline and revised text makes the comparison explicit. If commit was omitted, stop treating that chain as the intended control, start a safe copy again and retain the deviation note. Hosts may handle autocorrection, replacement and multicharacter input differently. The complete body after each action establishes the observed history boundary. Different step counts across hosts do not automatically make one wrong; the issue is whether observed states correspond to known targets. Keyboard appearance, skin name and font size do not determine edit-transaction granularity or guarantee an undo gesture in every field.

## 7. Diagnose missing controls, overshoot and neighbor changes

Without an Undo control, consult host help or system applicability and stop if unavailable rather than guessing a private feature from a long press. If disabled, check the same document, a genuinely committed edit and possible history loss after a session change. If Undo overshoots the baseline, restore the practice copy from the separately retained sentence and record the actual grouping; manual retyping is not successful Redo. If identifiers or punctuation also change, inspect the original selection and each state to distinguish initial overwriting from history behavior. Repair one locatable problem per trial while preserving references. Do not delete wider ranges in an official document to pursue an idealized one-character-per-action result.

## 8. Deliver an observed chain, not a fixed-step promise

Deliver baseline, revised sentence, Undo and Redo states in actual sequence, and notes on unavailable or unperformed stages. Identify the action producing each state and list manual restoration separately from system history. Finally choose the exercise state to retain, save through the host and reopen to compare text. This final step checks persistence only, not that history survives reopening. The outcome explains how this host grouped one particular revision, whether the two targets were reached and what remains unknown. It is not a universal undo specification. A different host, input method or autocorrection condition requires new observation; accurate entry of two digits with Jianlai does not establish full editing-history compatibility.

## Related guides and sources

[Local revision and neighbour check](<https://allluckly.cn/en/guides/jianlai-one-character-revision-neighbour-check/>); [One-setting baseline recovery](<https://allluckly.cn/en/guides/jianlai-baseline-one-setting-at-time-recovery/>).

Checked on 17 November 2026: [Apple’s Finland lookup](<https://itunes.apple.com/lookup?id=6754281557&country=fi&entity=software&timestamp=20261117>) and [China lookup](<https://itunes.apple.com/lookup?id=6754281557&country=cn&entity=software&timestamp=20261117>) both list Jianlai 1.0.5, requiring iOS / iPadOS 16.0 or later. The [official store](<https://apps.apple.com/fi/app/id6754281557>) describes themes, backgrounds, fonts and multilingual input. Use controls actually present in the installed version; system editing and host processing are distinguished separately. These are repeatable instructions and paper expectations, not claims of completed device input, host compatibility or format-parser tests.

Method references: [Apple text editing guide](<https://support.apple.com/guide/iphone/select-edit-and-move-text-iph1a9cae52c/ios>).

