You have a 50,000-line log file. You need all errors with their timestamps.
\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}.*ERROR.*
This grabs every line containing “ERROR” that starts with an ISO timestamp. With grep -P or your editor’s search, you go from 50,000 lines to the 12 that matter.
Level up: capture the error message:
\d{4}-\d{2}-\d{2}T[\d:]+.*ERROR\s+(.*)
Group 1 gives you just the error messages, stripped of timestamps and metadata.
Your API returns user IDs in different formats across endpoints:
"user_id": "USR-12345""userId": 67890user_id=USR-99999One regex catches all of them:
(?:user_?[Ii]d)["\s:=]+(?:"?)([A-Z]*-?\d+)
Database logs often include execution time:
Query completed in 1523ms: SELECT * FROM users WHERE...
Query completed in 45ms: SELECT id FROM products WHERE...
Query completed in 8920ms: SELECT * FROM orders JOIN...
Find queries over 1 second:
completed in (\d{4,})ms
\d{4,} matches 4 or more digits, meaning 1000ms or more.
Need to convert all console.log statements to use a logger?
Find:
console\.log\((['"`].*?['"`])\)
Replace with:
logger.info($1)
Most editors support regex find-and-replace. This handles thousands of instances in one operation.
You suspect some records in a CSV have malformed email addresses:
^[^,]+,(?![\w.+-]+@[\w-]+\.[\w.]+,)
Using a negative lookahead (?!...), this matches rows where the second column (assuming email) doesn’t match a valid email pattern.
Extract file names and line numbers from a Python stack trace:
File "([^"]+)", line (\d+)
Group 1: file path. Group 2: line number. Feed this into a script and you have an automated stack trace parser.
Scan your codebase for potential API keys and tokens:
(?:api[_-]?key|token|secret|password)\s*[=:]\s*["']([^"']+)["']
This isn’t foolproof, but it’s a solid first pass that catches the obvious ones.
Remove all HTML tags from text:
<[^>]+>
Strip multiple spaces to single:
\s{2,}
Replace with a single space.
Remove non-ASCII characters:
[^\x00-\x7F]
^ and $ prevent unexpected partial matches..*? instead of .* when you want the shortest match.Most debugging regex follows the same structure: timestamp/context pattern + separator + the thing you actually care about. Master this template and you can construct a useful pattern in under a minute.
]]>Stop thinking in local times. Think in UTC offsets.
When someone says “Let’s meet at 2pm,” the first question is: whose 2pm?
The overlap window between two time zones is where real-time collaboration happens. Everything else should be async.
India (IST) and US East (ET): 10.5 hour difference.
India and US West (PT): 13.5 hour difference.
India and UK (GMT): 5.5 hour difference.
If the meeting is always at a bad time for the same team, resentment builds. Alternate between time slots that inconvenience different groups.
If your overlap is 1.5 hours, protect it. Don’t waste it on status updates that could be async. Use synchronous time for decisions, brainstorming, and relationship-building.
“Meeting at 3pm” is ambiguous. “Meeting at 3pm ET / 12:30am IST” is not. Even better: share a link that converts to local time (Google Calendar does this automatically).
The US, Europe, and Australia change clocks on different dates. For 2-3 weeks per year, your carefully calculated offsets are wrong by an hour. Set calendar reminders around DST transitions.
Time zone overlap is precious and limited. Maximize it by making everything possible async:
Subtract 10.5 hours (subtract 11, add 30 min). IST 3pm = ET 4:30am.
Subtract 13.5 hours. IST 3pm = PT 1:30am.
Subtract 5.5 hours. IST 3pm = GMT 9:30am.
12:00 noon IST = 6:30am GMT = 1:30am ET = 10:30pm PT (previous day).
Memorize one anchor conversion for each zone pair you deal with regularly. Derive everything else from that.
The goal isn’t to make time zones disappear. It’s to make them a solved problem that doesn’t consume mental energy every day.
]]>RGB stands for Red, Green, Blue. Each channel ranges from 0 to 255.
rgb(255, 0, 0) = pure redrgb(0, 255, 0) = pure greenrgb(0, 0, 255) = pure bluergb(255, 255, 255) = white (all channels maxed)rgb(0, 0, 0) = black (all channels off)RGB maps directly to how screens work: each pixel has red, green, and blue sub-pixels that combine to produce color. It’s the native language of displays.
Good for: programmatic color manipulation, understanding how display hardware works, working with image data.
Bad for: intuiting what a color looks like from its values. Quick: is rgb(180, 120, 60) warm or cool? Hard to tell without experience.
HEX is just RGB written in hexadecimal. #FF0000 is rgb(255, 0, 0).
The six characters are three pairs: #RRGGBB. Each pair is a hex value from 00 (0) to FF (255).
#FF0000 = red (FF red, 00 green, 00 blue)#00FF00 = green#FFFFFF = white#000000 = blackShorthand: #F00 expands to #FF0000.
Good for: CSS (it’s the most common format), compact notation, copy-pasting between design tools and code.
Bad for: same as RGB, it’s hard to mentally manipulate. Want a slightly lighter version of #2563EB? Good luck doing that in your head.
HSL stands for Hue, Saturation, Lightness.
hsl(220, 83%, 53%) is a vivid blue. You can tell because:
Good for: designing color palettes, creating variations of a color, understanding color relationships.
Want five shades of your brand blue?
In HEX, you’d need to look up or calculate each one:
#1E40AF, #2563EB, #3B82F6, #60A5FA, #93C5FDIn HSL, just change lightness:
hsl(220, 83%, 35%) - darkhsl(220, 83%, 45%) - medium darkhsl(220, 83%, 53%) - basehsl(220, 83%, 65%) - medium lighthsl(220, 83%, 80%) - lightSame hue, same saturation, just sliding the lightness. Intuitive.
Want a muted version? Drop saturation. Want a complementary color? Add 180 to the hue.
| Format | Use when |
|---|---|
| HEX | Writing CSS, sharing colors, design specs |
| RGB | Working with canvas, image processing, programmatic color generation |
| HSL | Designing palettes, creating hover states, accessible color variations |
When choosing text and background colors, what matters is the lightness contrast. In HSL, this is immediately visible: if your background is hsl(220, 83%, 90%) and your text is hsl(220, 83%, 20%), the 70-point lightness gap tells you the contrast is strong.
WCAG requires a contrast ratio of at least 4.5:1 for normal text. Using HSL makes it much easier to intuitively maintain sufficient contrast.
]]>Base64 takes any data (text, images, files) and converts it into a string using only 64 characters: A-Z, a-z, 0-9, +, and /. Plus = for padding.
The input “Hello” becomes “SGVsbG8=”.
Every 3 bytes of input become 4 characters of output. This means Base64 encoded data is always about 33% larger than the original. It’s not compression. It’s the opposite.
Many systems were designed to handle text only. Email (SMTP) was designed for ASCII text. HTTP headers are text. JSON is text. XML is text. URLs are text.
When you need to send binary data (an image, a PDF, raw bytes) through these text-only channels, you have a problem. Binary data contains bytes that can be misinterpreted as control characters, get corrupted by text processing, or simply aren’t valid in the format.
Base64 solves this by encoding binary data into safe text characters that survive any text-based transport.
When you attach a photo to an email, it’s Base64 encoded into the MIME message. Your email client decodes it back to the original image.
<img src="data:image/png;base64,iVBORw0KGgo..." />
Small images can be embedded directly in HTML without separate HTTP requests.
HTTP Basic Auth sends credentials as username:password Base64 encoded in the Authorization header.
Important: Base64 is NOT encryption. Anyone can decode it. Basic Auth is only safe over HTTPS.
JSON has no binary type. If you need to include file data in a JSON payload, Base64 is the standard approach.
JSON Web Tokens are three Base64-encoded segments separated by dots. The payload is readable by anyone who decodes it (again, it’s not encryption).
+ and /. Padding with =.- and _ instead (because + and / have special meaning in URLs).= signs. Some systems don’t need them.Base64 is a tool for making binary data survive text-based transport. It’s not security. It’s not compression. It makes things bigger but compatible. Use it when you must, avoid it when you can send binary directly.
]]>{
"name": "Amit",
"age": 30,
"is_developer": true,
"skills": ["Python", "JavaScript", "SQL"],
"address": {
"city": "Bangalore",
"country": "India"
}
}
That’s it. If you can read the above, you can read JSON.
JSON has six data types:
"hello"42 or 3.14true or false (lowercase, no quotes)null (represents nothing)[1, 2, 3]{"key": "value"}And a few syntax rules:
Before JSON, XML was the standard data format:
<person>
<name>Amit</name>
<age>30</age>
</person>
Compare to JSON:
{"name": "Amit", "age": 30}
JSON is shorter, easier to read, easier to parse, and maps directly to data structures in most programming languages. XML has more features (schemas, namespaces, attributes), but for 90% of use cases, JSON’s simplicity wins.
package.json, tsconfig.json, VS Code settings, all JSON.{
"a": 1,
"b": 2,
}
That trailing comma after 2 is invalid. JSON parsers will reject it.
{'name': 'Amit'}
Invalid. JSON requires double quotes.
{name: "Amit"}
Invalid. Keys must be quoted strings.
{
"name": "Amit" // this is a comment
}
Invalid. JSON doesn’t support comments. If you need comments in config files, consider JSONC or YAML instead.
Paste your JSON into a formatter/validator tool. It will instantly tell you if it’s valid and show you where the error is. When debugging API responses or config files, this is the first thing to do.
JSON.parse() in JavaScript, json.loads() in Python, json.Unmarshal() in Go.| Pattern | Matches |
|---|---|
. |
Any single character |
\d |
Any digit (0-9) |
\w |
Any word character (a-z, A-Z, 0-9, _) |
\s |
Any whitespace (space, tab, newline) |
^ |
Start of string |
$ |
End of string |
* |
Zero or more of previous |
+ |
One or more of previous |
? |
Zero or one of previous |
{n,m} |
Between n and m of previous |
[abc] |
Any character in the set |
[^abc] |
Any character NOT in the set |
(...) |
Capture group |
\| |
OR operator |
[\w.+-]+@[\w-]+\.[\w.]+
Not RFC-compliant (nothing is, practically), but catches 99% of real emails.
https?://[\w\-.]+(:\d+)?(/[\w\-./?%&=]*)?
(\+91[\s-]?)?[6-9]\d{9}
(\+?1[\s-]?)?\(?\d{3}\)?[\s-]?\d{3}[\s-]?\d{4}
\b\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}\b
\d{4}-(?:0[1-9]|1[0-2])-(?:0[1-9]|[12]\d|3[01])
(?:0[1-9]|[12]\d|3[01])/(?:0[1-9]|1[0-2])/\d{4}
(?:[01]\d|2[0-3]):[0-5]\d(:[0-5]\d)?
-?\d+
-?\d+\.?\d*
#([0-9a-fA-F]{3}|[0-9a-fA-F]{6})\b
</?[\w]+[^>]*>
^\s+|\s+$
Use with replace to trim leading/trailing whitespace.
\b(\w+)\s+\1\b
Finds “the the”, “is is”, etc.
[A-Z]{5}\d{4}[A-Z]
\b[1-9]\d{5}\b
\b\d{5}(-\d{4})?\b
^(?=.*[a-z])(?=.*[A-Z])(?=.*\d).{8,}$
https?://([^/]+)
Capture group 1 gives you the domain.
(?:^|,)("(?:[^"]*"")*[^"]*"|[^,]*)
(?:...) when you don’t need to extract the match.\d{4} is better than \d+ when you know exactly 4 digits are expected.string.contains() or string.startsWith() would do.Regex is a scalpel. Use it when you need precision. Don’t use it when a butter knife works fine.
]]>The Institute of Medicine suggests:
“Total water” includes water from food (which accounts for roughly 20-30% of intake). So the actual drinking requirement is lower.
But these are averages for temperate climates with moderate activity. Your needs vary based on several factors.
Larger bodies need more water. A rough formula: 30-35 ml per kg of body weight.
A 70 kg person: 2.1 to 2.45 liters. A 90 kg person: 2.7 to 3.15 liters.
You lose 0.5 to 1 liter of water per hour of moderate exercise through sweat. Intense exercise in heat can push this to 2+ liters per hour.
Hot and humid weather increases sweat losses. Dry climates increase water loss through respiration and skin. Both require extra intake.
High-sodium food increases water needs. High-water-content foods (fruits, vegetables, soups) reduce how much you need to drink separately. A person eating lots of watermelon and cucumber needs less drinking water than someone eating mostly dry, processed food.
Above 2,500 meters, your body loses more water through increased respiration and urine output. High-altitude trekkers should increase intake by 0.5-1 liter.
For a sedentary person in a temperate climate eating a balanced diet, 8 glasses (about 2 liters) is… actually fine. Not because the rule is scientifically derived, but because it happens to land in the right ballpark for many people. It’s an okay default, just not a universal truth.
Forget counting glasses. Your body has a built-in hydration meter.
Check first thing in the morning. That’s your baseline hydration level.
Yes. Hyponatremia (dangerously low blood sodium from excessive water intake) is rare but real. It most commonly affects endurance athletes who drink far more than they sweat. For most people, the kidneys handle excess water efficiently, but there’s no benefit to forcing down 5+ liters when your body doesn’t need it.
Protein: 4 calories per gram. Builds and repairs muscle, keeps you full, has the highest thermic effect (your body burns ~25% of protein calories just digesting it).
Carbohydrates: 4 calories per gram. Your body’s preferred energy source, especially for high-intensity activity. Brain fuel. Stored as glycogen in muscles and liver.
Fat: 9 calories per gram. Essential for hormones, brain function, cell membranes, and absorbing fat-soluble vitamins (A, D, E, K). More calorie-dense, so portions matter.
Protein is the most important macro to get right, regardless of your goal. The research is clear:
For a 75 kg person building muscle: 120 to 165 grams of protein per day.
Always set protein first, then distribute remaining calories between carbs and fat.
For our 75 kg person eating 2,100 cal/day:
Ratio: roughly 30% protein, 28% fat, 42% carbs.
At 3,000 cal/day:
Ratio: roughly 18% protein, 21% fat, 61% carbs.
| Macro | Good sources |
|---|---|
| Protein | Chicken, eggs, paneer, dal, Greek yogurt, fish, tofu, whey |
| Carbs | Rice, roti, oats, potatoes, fruits, vegetables |
| Fat | Nuts, ghee, olive oil, avocado, cheese, fatty fish |
Low-carb diets work for fat loss because cutting carbs often means cutting calories. There’s no metabolic magic. At equal calorie and protein levels, high-carb and low-carb diets produce similar fat loss results.
The best carb level is the one you can sustain. If you train intensely, you’ll perform better with moderate to high carbs. If you’re sedentary and find it easier to control hunger on lower carbs, that works too.
Perfecting macro ratios while eating the wrong total calories is like optimizing tire pressure on a car with no engine. Get the calories right first, then refine.
]]>Your TDEE is made up of four parts:
BMR (Basal Metabolic Rate): ~60-70% of total. This is what your body burns just existing. Breathing, circulating blood, maintaining body temperature. Even in a coma, you’d burn this much.
TEF (Thermic Effect of Food): ~10%. The energy cost of digesting food. Protein costs more to digest (~25%) than carbs (~8%) or fat (~3%).
EAT (Exercise Activity Thermogenesis): ~5-10%. Intentional exercise. The gym, running, sports.
NEAT (Non-Exercise Activity Thermogenesis): ~15-20%. Everything else. Walking to the kitchen, fidgeting, standing, typing. This is the most variable and underrated component.
The Mifflin-St Jeor equation is the most accurate for most people:
Men: BMR = 10 x weight(kg) + 6.25 x height(cm) - 5 x age - 5
Women: BMR = 10 x weight(kg) + 6.25 x height(cm) - 5 x age - 161
Example: 30-year-old male, 75 kg, 175 cm BMR = 750 + 1094 - 150 + 5 = 1,699 calories
Multiply BMR by an activity factor:
| Activity Level | Multiplier |
|---|---|
| Sedentary (desk job, no exercise) | 1.2 |
| Lightly active (1-3 days exercise/week) | 1.375 |
| Moderately active (3-5 days/week) | 1.55 |
| Very active (6-7 days/week) | 1.725 |
| Extremely active (physical job + training) | 1.9 |
Our example person, moderately active: 1,699 x 1.55 = 2,633 calories/day
Eat below TDEE. A deficit of 500 calories/day produces roughly 0.5 kg of fat loss per week (since 1 kg of fat = ~7,700 calories).
For our example: 2,633 - 500 = 2,133 calories/day for steady fat loss.
Eat above TDEE. A surplus of 300-500 calories supports muscle growth without excessive fat gain.
Target: 2,633 + 400 = 3,033 calories/day with strength training.
Eat at TDEE. Weight stays stable.
Most people with desk jobs who exercise 3 times a week are “lightly active,” not “moderately active.” When in doubt, pick the lower multiplier.
NEAT can vary by 500-2,000 calories between individuals. Someone who walks 12,000 steps daily and fidgets constantly burns dramatically more than someone who sits still all day. Increasing NEAT (walking more, standing desk, taking stairs) is often easier than adding gym sessions.
As you lose weight, your BMR drops. A person who lost 10 kg needs to recalculate. This is why weight loss plateaus happen.
TDEE calculations are estimates with a margin of error around 10-15%. Use the number as a starting point, track your weight for 2-3 weeks, and adjust based on actual results.
Body Mass Index = weight (kg) / height (m) squared.
It was invented in the 1830s by a Belgian mathematician named Adolphe Quetelet. He wasn’t a doctor. He was studying population statistics. BMI was designed to categorize populations, not diagnose individuals.
For the average person who doesn’t exercise intensely, BMI is a reasonable proxy for body fat. Large population studies consistently show that BMI correlates with health risks:
For a doctor screening thousands of patients, BMI is quick, cheap, and “good enough” to flag who needs closer examination.
A muscular person and an overweight person at the same height and weight have identical BMIs. Dwayne “The Rock” Johnson’s BMI is around 34 (obese). He’s clearly not.
Where you carry fat matters more than how much. Visceral fat around your organs is far more dangerous than subcutaneous fat under your skin. Two people with BMI 27 can have very different health profiles.
The standard BMI categories were developed using data from European populations. For South Asians, health risks rise at lower BMI values. Many researchers suggest using 23 (not 25) as the overweight threshold for Asian populations.
BMI doesn’t adjust for the fact that body composition naturally changes with age, or that women carry more essential fat than men.
The single best addition to BMI. Measure around your navel.
Waist measurement divided by hip measurement.
Your waist should be less than half your height. Simple, effective, works across populations.
The gold standard if you can measure it. Healthy ranges:
DEXA scans and bioimpedance scales can estimate this, with varying accuracy.
Don’t ignore BMI entirely. It’s free, instant, and gives you a starting point. But don’t let a number from an 1830s Belgian statistician be your only health metric. Pair it with waist circumference at minimum. If your BMI says you’re fine but your waist says otherwise, trust the waist.
]]>