René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Inconsistent result of failed Base64.decode (throws vs returns null)","articleBody":"When `Base64.decode(CharSequence in)` decodes a string, in some cases for \"bad\" strings in throws `IllegalArgumentException`:\r\n\r\nhttps://github.com/patrickfav/bytes-java/blob/e622b3b4dfe0bca2b65d2dcd6dae78441ee2961a/src/main/java/at/favre/lib/bytes/Base64.java#L92\r\n\r\n, but in other cases it returns `null`.\r\n\r\nhttps://github.com/patrickfav/bytes-java/blob/e622b3b4dfe0bca2b65d2dcd6dae78441ee2961a/src/main/java/at/favre/lib/bytes/Base64.java#L108-L110\r\n\r\nThis makes the method harder to use as the caller needs to handle both cases for every call with potentially invalid input: catching `IllegalArgumentException` and comparing the result with `null`.\r\nI think the best way to handle both types of invalid input would be to throw `IllegalArgumentException` in both cases and never return `null`.\r\n\r\nThis behavior is even tested:\r\n\r\nhttps://github.com/patrickfav/bytes-java/blob/e622b3b4dfe0bca2b65d2dcd6dae78441ee2961a/src/test/java/at/favre/lib/bytes/Base64Test.java#L47\r\n\r\nIt makes me think that this is the desired behavior, but why?\r\n\r\nBtw, this behavior actually leads to the bug in the `id-mask` library. That is how I actually found it.\r\nWhen I do `idMask.unmask(badlyEncodedString)`, then it eventually calls `Base64.decode`, that returns `null`, and then it tries to wrap it into `Bytes` object\r\n\r\nhttps://github.com/patrickfav/bytes-java/blob/e622b3b4dfe0bca2b65d2dcd6dae78441ee2961a/src/main/java/at/favre/lib/bytes/Bytes.java#L155-L157\r\n\r\n, that leads to `java.lang.NullPointerException: passed array must not be null`. That is unexpected behavior I believe. `idMask.unmask(badlyEncodedString)` should never throw NPE.\r\n\r\nOf course, this problem could be solved on the `id-mask` side (by catching NPE and throwing `IllegalArgumentException`), but I think the root problem is the inconsistent behavior of `Base64.decode`. I can't imagine a situation when it is the desired behavior to sometimes throw `IllegalArgumentException` and sometimes return `null` for basically the same case of \"invalid input\".\r\n\r\nWhy it is important to throw `IllegalArgumentException` and not `NullPointerException`? Because the 1st error means the error is on the caller side (probably someone called the API with a wrong encoded ID), while the 2nd one means that something wrong happened inside the library. NPE would probably be acceptable in the case when `input` CharSequence is `null`. But if it is a non-null string, then NPE should never be thrown except for a bug in the implementation.","author":{"url":"https://github.com/xak2000","@type":"Person","name":"xak2000"},"datePublished":"2024-03-20T17:53:12.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":0},"url":"https://github.com/65/bytes-java/issues/65"}