Regex finds comma wenn searching for period

[KasimirNepomuk]

Hey there,

I wanted to use regexReplacer in the String manipulation node to remove periods in numbers, so for example: 1.000 → 1000

I used the following regex for this operation: regexReplace($Value$,“([0-9])\.([0-9])” ,“$1$2” )

Now, according to https://regex101.com/, this should yield the correct values, and it does. However, it also removes commas: 3,5 → 35

This should not be the case. I have attached the workflow. I work on Windows 11. My KNIME version should be up to date: KNIME 5.7.0.

RegexCommaBug.knwf (44.5 KB)

[mlauber71]

@KasimirNepomuk I cannot test it right now but you should try and double escape the dot.

[KasimirNepomuk]

Hello, thank you, this seams to work. But why does it need to be doubly escaped and why does it only remove commas?

The last question I was able to answer myself: there are no other values except for the comma and the period that are between two numbers and not part of another group.

[takbb]

Hi @KasimirNepomuk , the “double escaping” is generally needed in KNIME (and some other languages) when the regex is in a literal string within a script. This is because KNIME first has to interpret the string, which can also contain escapes. So in the first level of parsing, the single `\` is treated as an escape within the script, and this is before it is passed to the regex engine, so what gets passed to regex has had a single escape already stripped.

You can find a little more detail on this here

How to replace or remove the \ character?

Hi @Thoralf , you are right that the forum software gives special meaning to some characters, and if pasting certain text (especially if you want it to appear exactly as written) you need to highlight it and use the “preformatted text” button on the forum toolbar: [image]. This also prevents it doing unwanted conversions of "double quotes" into “smart quotes”. In the same way, regex, along with some other scripting languages, gives special meaning to the \ character so if you want it to be t…

[gonhaddock]

Hello The provided WF seems to be empty for me. You may not need a regexReplace(), you can test with standard replace() in a ‘String Manipulation’ node:

replace($Value$, ".", "")

Regarding double scape… some nodes require it aiming to translate from standard regex syntax to supported Java syntax like in ‘String Manipulation’ or ‘Expression’ nodes, then double scape is needed:

This syntax seems to literal for me and it has many fail cases:

regex_replace($["Value"], "^(\\d+)\\.(\\d+)", "$1$2") (‘Expression’ node)

regexReplace($["Value"], "^([0-9]+)\\.([0-9]+)", "$1$2") (‘String Manipulation’ node)

This expression would be valid as well, and it’s suitable if you have consecutive thousand groups:

regex_replace($["Value"], "(\\d+[,]?)[.]?", "$1") (‘Expression’ node)

regexReplace($["Value"], "(\\d+[,]?)[.]?", "$1") (‘String Manipulation’ node)

BR