CSV Number Format Converter

A CSV Number Format Converter prepares explicitly numeric fields for a receiving application without assuming that every digit sequence is a quantity. Prices, weights, and percentages often arrive with incompatible separators. Customer identifiers, postal codes, phone numbers, and long order references can resemble numbers while requiring exact text preservation. This tool initially selects no columns. Add a conversion only for a field whose numeric meaning and source notation you have established.

Conversion creates a new numeric-text column and a companion status column. The original input is retained, including leading zeros, symbols, and spacing. Parsed values are stored separately in the interpretation report. This separation lets you compare a supplier's original expression with the value used for calculation and the final output spelling. The converter does not exchange currencies, infer a currency from a country, or strip all nonnumeric characters to make a cell appear valid.

Choose files or Paste a table and confirm the CSV delimiter, quoting, and header. A comma serving as a decimal separator is different from a comma separating CSV fields. A cell containing 1.234,50 therefore needs proper CSV quoting when the file delimiter is a comma. Import parsing establishes cell boundaries before numeric interpretation begins. Changing the numeric settings cannot repair a file that was parsed into the wrong columns; revisit the import settings if the grid shows broken records.

Use Add conversion column and select Column to convert. Source decimal separator accepts a dot or comma. Source grouping separator can be absent, a dot, a comma, a regular space, a nonbreaking space, a narrow nonbreaking space, or an apostrophe. The two separators cannot be the same. Grouped integer sections must have a valid three-digit grouping pattern; a malformed group such as 12,34 is not silently flattened into 1234 when comma grouping is selected.

Negative notation distinguishes a leading minus sign from accounting parentheses. Allow both only if the source uses both forms. A leading plus sign is supported, but contradictory signs or mixed parentheses and signs are rejected. Allowed currency tokens lists literal prefixes or suffixes that may be removed. Undeclared units, explanatory text, or a currency symbol not on that list remain invalid. Adding a token is permission to remove that exact marker, not permission to infer its economic meaning.

Whitespace inside a number is accepted as a grouping separator only when that exact character is configured. A regular space and a nonbreaking space are different characters. Surrounding spaces are not automatically trimmed by the numeric parser. If those spaces are unwanted source noise, use an explicit cleaning step before conversion and keep that step in the workflow. Empty strings are separate from zero. Literal NULL, N/A, and false are not missing unless you deliberately add them to Empty definition.

Percentage interpretation defaults to rejecting percent signs. Percent value interprets 12% as the numeric value 12, while Ratio interprets it as 0.12. Choose the form expected by the receiving system. A proportion column and a percentage-points column can both contain useful data, but mixing their values can create a hundredfold error. The interface language never chooses the percentage mode or the decimal convention for you.

For a practical example, take two fields called order_id and discount. Keep order_id untouched even if its value is 000123. Add a conversion for discount and choose Ratio. An input of 12% then produces 0.12 in the new result column while the original remains 12%. If a second record contains a bare 12 without a percent sign, it remains the number 12 under the configured numeric convention. The presence of other percentage cells does not authorize attaching a missing percent sign.

Output decimal separator and Output grouping separator control the spelling of successful results. With source decimal comma and grouping dot, 1.234,50 becomes 1234.50 when the output uses a dot and no grouping. Keep trailing zeros preserves the source decimal scale unless you request fixed decimal places. Without trailing zeros, the same numeric value can be written as 1234.5. Neither choice changes the exact parsed value used by other calculations.

Leave Fixed decimal places blank to avoid rounding. If you enter a number of places, choose a Rounding mode and confirm that fixed precision may change values. Nearest with ties away from zero turns 1.005 into 1.01 at two decimal places and turns -1.005 into -1.01. Nearest with ties to even produces 1.00 and -1.00 for those boundary inputs. Toward zero, toward negative infinity, and toward positive infinity are separate policies, especially for negative values.

The shared numeric path uses decimal arithmetic rather than binary floating-point addition for supported quantities. Consequently, an exact sum of 0.1 and 0.2 is 0.3 in the grouping and pivot tools. Long integers are interpreted without passing through a JavaScript binary-number conversion. A value such as 100000000000000000000000000000.01 retains its decimal meaning within the supported input bound. An unselected identifier preserves its complete original spelling rather than entering arithmetic at all.

Numeric interpretation is bounded to one thousand input digits and controlled output precision. Scientific notation is not accepted by this source parser; expand it to explicit decimal text using a trusted source system if necessary. Oversized values receive unsupported status rather than a plausible rounded approximation. For statistics, recurring decimal means are reported to at most forty fractional places under the documented decimal rounding policy. That averaging policy does not authorize loss of significant digits during ordinary format conversion.

Confirm this numeric interpretation records that the source convention was reviewed. Run and review then produces successful, empty, invalid, ambiguous where applicable, and unsupported result views. Numeric conversion reports also show whether fixed precision rounded a value and whether its output spelling changed. An unchanged value is a valid outcome. If the input already matches the requested representation, there is no need to invent a visible modification to demonstrate that processing occurred.

Inspect the interpretation record for any failed value. It includes the source record, original text, parsed value when available, output, and reason. A failed conversion writes a blank result with a failure status while retaining the original field. It never turns failure into zero. This distinction matters when the output will feed a sum: an actual zero contributes a valid observation, an empty cell contributes no number, and an invalid amount must be resolved or explicitly excluded under a reported policy.

Export new copy lets you choose the complete result, a status report, or a bundle. Review the number of records and fields, CSV or TSV delimiter, UTF-8 encoding, and formula-prefix protection before downloading. Protection may add an apostrophe to legitimate negative numbers or plus-prefixed phone values; its effect is counted. Choose original text only after acknowledging how spreadsheet applications may interpret it. Read-back verification checks generated file structure and values, but cannot control another application's automatic type inference.

Apply to workflow passes the result and its confirmed numeric interpretation to later validation, grouping, and pivot operations. A normalized result column carries its output separators as the next source convention, avoiding a second guess based on interface language. Save rules preserves conversion choices, while Load rules checks the new file's field structure. Imported rule files can contain private field names and literal markers, so inspect them before sharing. Cancel task leaves source data available; editing settings makes old results stale until another run succeeds.

When most rows fail, first inspect the chosen source separators and the parsed cell boundaries. When only a few fail, examine their exact text for undeclared units, a different space character, malformed grouping, or a percentage sign. Do not keep broadening token removal merely to reduce the error count. When the file mixes several numeric conventions in one column, separate the records using reliable source information and apply the appropriate confirmed convention to each subset.

The sample library contains a normal decimal-comma example, a percentage and long-identifier boundary case, and undeclared-unit errors. Use these to learn the controls before loading sensitive data. Processing scope can restrict an exploratory run to the first twenty records, but its reports are explicitly labeled as samples and cannot be applied to the complete workflow. A successful sample says nothing about later records. Full processing and export remain subject to input, audit, output, and time limits rather than an unlimited-size promise.