Provide implementations of advanced data structures
中文简介正在准备中。当前显示原文说明。
Configure cli-progress with custom formatters, multi-bar support, and ETA calculations for CLI progress indication.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Configure cli-progress for advanced progress indication.
Invoke this skill when you need to:
| Parameter | Type | Required | Description |
|---|---|---|---|
| language | string | Yes | Target language |
| format | string | No | Bar format template |
| presets | array | No | Progress bar presets |
import cliProgress, { SingleBar, MultiBar, Presets } from 'cli-progress';
import chalk from 'chalk';
// Custom format with colors
const defaultFormat = chalk.cyan('{bar}') +
' {percentage}% | ETA: {eta}s | {value}/{total} | {filename}';
// Create single progress bar
export function createProgressBar(options?: {
format?: string;
barSize?: number;
}): SingleBar {
return new SingleBar({
format: options?.format || defaultFormat,
barCompleteChar: '\u2588',
barIncompleteChar: '\u2591',
barsize: options?.barSize || 40,
hideCursor: true,
clearOnComplete: false,
stopOnComplete: true,
}, Presets.shades_classic);
}
// Multi-bar for parallel tasks
export function createMultiBar(): MultiBar {
return new MultiBar({
format: '{name} |' + chalk.cyan('{bar}') + '| {percentage}% | {value}/{total}',
barCompleteChar: '\u2588',
barIncompleteChar: '\u2591',
hideCursor: true,
clearOnComplete: false,
stopOnComplete: false,
}, Presets.shades_grey);
}
// Download progress with speed
export function createDownloadBar(): SingleBar {
return new SingleBar({
format: 'Downloading |' + chalk.cyan('{bar}') +
'| {percentage}% | {speed} MB/s | {value}/{total} MB',
barCompleteChar: '\u2588',
barIncompleteChar: '\u2591',
barsize: 30,
formatValue: (value, options, type) => {
if (type === 'value' || type === 'total') {
return (value / 1024 / 1024).toFixed(2);
}
return String(value);
},
}, Presets.shades_classic);
}
// Wrapper for async operations
export async function withProgress<T>(
total: number,
fn: (update: (current: number, payload?: Record<string, any>) => void) => Promise<T>,
options?: { format?: string }
): Promise<T> {
const bar = createProgressBar(options);
bar.start(total, 0);
try {
const result = await fn((current, payload) => {
bar.update(current, payload);
});
bar.stop();
return result;
} catch (error) {
bar.stop();
throw error;
}
}
// Batch processing with progress
export async function processBatch<T, R>(
items: T[],
processor: (item: T, index: number) => Promise<R>,
options?: { label?: string }
): Promise<R[]> {
const bar = createProgressBar({
format: `${options?.label || 'Processing'} |{bar}| {percentage}% | {value}/{total}`,
});
bar.start(items.length, 0);
const results: R[] = [];
for (let i = 0; i < items.length; i++) {
results.push(await processor(items[i], i));
bar.update(i + 1);
}
bar.stop();
return results;
}
// Parallel processing with multi-bar
export async function processParallel<T, R>(
tasks: Array<{
name: string;
items: T[];
processor: (item: T) => Promise<R>;
}>
): Promise<Map<string, R[]>> {
const multiBar = createMultiBar();
const bars = new Map<string, SingleBar>();
const results = new Map<string, R[]>();
// Create bars for each task
for (const task of tasks) {
const bar = multiBar.create(task.items.length, 0, { name: task.name.padEnd(15) });
bars.set(task.name, bar);
results.set(task.name, []);
}
// Process all tasks in parallel
await Promise.all(tasks.map(async (task) => {
const bar = bars.get(task.name)!;
const taskResults = results.get(task.name)!;
for (let i = 0; i < task.items.length; i++) {
taskResults.push(await task.processor(task.items[i]));
bar.update(i + 1);
}
}));
multiBar.stop();
return results;
}
{
"dependencies": {
"cli-progress": "^3.12.0",
"chalk": "^5.0.0"
},
"devDependencies": {
"@types/cli-progress": "^3.11.0"
}
}
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Provide implementations of advanced data structures
中文简介正在准备中。当前显示原文说明。
Fresh adversarial code review with binary PASS/FAIL verdicts, evidence citations, and anchoring bias prevention via fresh reviewer spawning.
中文简介正在准备中。当前显示原文说明。
WASM-based instant code transforms for simple tasks, achieving 352x speedup over LLM inference with zero cost.
中文简介正在准备中。当前显示原文说明。
Coordinate Crew (persistent) and Polecat (transient) agents using Gas Town's hook-based work distribution and GUPP principle.
中文简介正在准备中。当前显示原文说明。
Generate visual representations of algorithm execution
中文简介正在准备中。当前显示原文说明。
Hierarchical coordination and drift detection with frequent checkpoints, shared memory coherence validation, role specialization enforcement, and short task cycles.
中文简介正在准备中。当前显示原文说明。