You will often see developers use a special sentinel for unbounded ranges, priority queues, or shortest-path algorithms. Python makes this straightforward, but the details can trip you up when you mix in JSON, Decimal, NumPy, or sorting edge cases. <\/p>\n\n\n\n
How to represent infinity in Python, both positive and negative? What are the practical differences between using float(‘inf’) and math.inf? How do I check for it, sort with it, serialize it to JSON, and work with it in Decimal or NumPy without surprises?<\/em><\/p>\n\n\n\n The canonical way to represent infinity in Python uses a special floating point value Infinity is useful when sorting values where some entries are missing or should sort to the end.<\/p>\n\n\n If there’s no boundary, use infinity in your function parameters or algorithms.<\/p>\n\n\n NumPy has numpy.inf and compatible checks.<\/p>\n\n\n The JSON standard does not include infinity or NaN. Python\u2019s If untrusted input can flow into numeric code, block infinities and In Python image or video pipelines, you might treat a missing upper bound as infinity and then clamp before generating delivery URLs. This avoids sending invalid values to APIs and ensures safe transforms.<\/p>\n\n\nAnswer:<\/h2>\n\n\n\n
float(\u201cinf)<\/code>. You can create positive and negative infinity like this:<\/p>\n\n\nimport math\n\npos_inf = float('inf'<\/span>)\u00a0 \u00a0 \u00a0 \u00a0 # Positive infinity<\/span>\nneg_inf = float('-inf'<\/span>) \u00a0 \u00a0 \u00a0 # Negative infinity<\/span>\nalso_inf = math.inf \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # Same as float('inf')<\/span>\n\nprint<\/span>(pos_inf == also_inf)\u00a0 \u00a0 # True<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nHow infinity behaves<\/h3>\n\n\n\n
\n
math.isinf(x)<\/code>, math.isfinite(x)<\/code>, math.isnan(x)<\/code>.<\/li>\n<\/ul>\n\n\nimport math\n\nx = float('inf'<\/span>)\nprint<\/span>(x > 10<\/span>**30<\/span>) \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # True<\/span>\nprint<\/span>(math.isinf(x))\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # True<\/span>\nprint<\/span>(math.isfinite(x)) \u00a0 \u00a0 \u00a0 # False<\/span>\nprint<\/span>((x - x))\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # nan<\/span>\nprint<\/span>(math.isnan(x - x))\u00a0 \u00a0 \u00a0 # True<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nSorting and sentinels<\/h3>\n\n\n\n
data = [3<\/span>, None, 7<\/span>, -2<\/span>, None]\n\n# Move None to the end by substituting +inf for the sort key<\/span>\nsorted_data = sorted(data, key=lambda v: v if<\/span> v is not None else<\/span> float('inf'<\/span>))\nprint<\/span>(sorted_data)\u00a0 # [-2, 3, 7, None, None]<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nClamping and unbounded defaults<\/h3>\n\n\n\n
def clamp(x, low=-float('inf'<\/span>), high=float('inf'<\/span>)):\n\u00a0 \u00a0 if<\/span> x < low:\n\u00a0 \u00a0 \u00a0 \u00a0 return<\/span> low\n\u00a0 \u00a0 if<\/span> x > high:\n\u00a0 \u00a0 \u00a0 \u00a0 return<\/span> high\n\u00a0 \u00a0 return<\/span> x\n\nprint<\/span>(clamp(10<\/span>, high=5<\/span>))\u00a0 \u00a0 \u00a0 # 5<\/span>\nprint<\/span>(clamp(10<\/span>))\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # 10\u00a0 (no bounds)<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nDecimal and Fraction<\/h3>\n\n\n\n
\n
from decimal import Decimal\n\nd_inf = Decimal('Infinity'<\/span>)\nd_ninf = Decimal('-Infinity'<\/span>)\nprint<\/span>(d_inf > Decimal('1e999'<\/span>))\u00a0 # True<\/span>\n\n# Fractions do not have an infinity representation<\/span>\n# Fraction(1, 0) would raise ZeroDivisionError<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nNumPy<\/h3>\n\n\n\n
import numpy as<\/span> np\n\narr = np.array<\/span>([1.0<\/span>, np.inf, -np.inf, np.nan])\nprint<\/span>(np.isinf(arr))\u00a0 \u00a0 \u00a0 # [False\u00a0 True\u00a0 True False]<\/span>\nprint<\/span>(np.isfinite(arr)) \u00a0 # [ True False False False]<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nJSON and persistence pitfalls<\/h3>\n\n\n\n
json.dumps<\/code> will emit non-standard Infinity<\/code> by default, which some parsers reject. Set allow_nan=False<\/code> or sanitize values.<\/p>\n\n\nimport json, math\n\npayload = {\"max\"<\/span>: float('inf'<\/span>)}\n\nprint<\/span>(json.dumps(payload)) \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 # {\"max\": Infinity}\u00a0 not strict JSON<\/span>\ntry<\/span>:\n\u00a0 \u00a0 json.dumps(payload, allow_nan=False<\/span>) \u00a0 # raises ValueError<\/span>\nexcept ValueError:\n\u00a0 \u00a0 pass\n\ndef sanitize_numbers(x):\n\u00a0 \u00a0 if<\/span> isinstance(x, float) and<\/span> (math.isinf(x) or<\/span> math.isnan(x)):\n\u00a0 \u00a0 \u00a0 \u00a0 return<\/span> None\n\u00a0 \u00a0 if<\/span> isinstance(x, dict):\n\u00a0 \u00a0 \u00a0 \u00a0 return<\/span> {k: sanitize_numbers(v) for<\/span> k, v in x.items()}\n\u00a0 \u00a0 if<\/span> isinstance(x, list<\/span>):\n\u00a0 \u00a0 \u00a0 \u00a0 return<\/span> [sanitize_numbers(v) for<\/span> v in x]\n\u00a0 \u00a0 return<\/span> x\n\nprint<\/span>(json.dumps(sanitize_numbers(payload)))\u00a0 # {\"max\": null}<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nGuardrails for user input<\/h3>\n\n\n\n
NaNs<\/code> to avoid undefined math.:<\/p>\n\n\ndef ensure_finite(x: float) -> float:\n\u00a0 \u00a0 if<\/span> not math.isfinite(x):\n\u00a0 \u00a0 \u00a0 raise ValueError(\"Expected a finite number\"<\/span>)\n\u00a0 \u00a0 return<\/span> x<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\nUsing infinity safely with Cloudinary workflows<\/h3>\n\n\n\n
import math\nfrom cloudinary.utils import cloudinary_url\n\ndef build_limited_url(public_id: str, target_w: float):\n\u00a0 \u00a0 # Treat infinity as \"no limit,\" but hard cap to a safe maximum<\/span>\n\u00a0 \u00a0 MAX_W = 4096<\/span>\n\u00a0 \u00a0 width = MAX_W if<\/span> not math.isfinite(target_w) else<\/span> int(min(target_w, MAX_W))\n\u00a0 \u00a0 url, _ = cloudinary_url(public_id, transformation=[{\"crop\"<\/span>: \"limit\"<\/span>, \"width\"<\/span>: width}])\n\u00a0 \u00a0 return<\/span> url\n\nprint<\/span>(build_limited_url(\"sample.jpg\"<\/span>, float('inf'<\/span>))) \u00a0 # uses width=4096<\/span>\nprint<\/span>(build_limited_url(\"sample.jpg\"<\/span>, 1200.0<\/span>)) \u00a0 \u00a0 \u00a0 \u00a0 # uses width=1200<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\nTL;DR<\/h3>\n\n\n\n
\n
float('inf')<\/code>, float('-inf')<\/code>, or math.inf<\/code>.<\/li>\n\n\n\nmath.isinf<\/code> and math.isfinite<\/code>. Beware that inf - inf<\/code> yields NaN.<\/li>\n\n\n\nLearn More<\/h2>\n\n\n\n