BOM Import Check for Japan: 359,135 Walsin Part Numbers Collapse to One Key Each, and 14 Non-ASCII Characters Sit in 1,061,402 Cells

A BOM import check for Japan has one job the export makes easy: every Walsin part number is plain ASCII, so any full-width character in your BOM is a guaranteed miss. We tested all 359,135 distinct part numbers: none matches as typed full-width, and one normalised key resolves every one of them with zero collisions.

Venatech Limited, an authorised Walsin distributor in Hong Kong, holds all 203 parametric export files published by Walsin Technology (Walsin). This post is written for engineers and buyers in Japan whose BOM has passed through a Japanese-language spreadsheet or input method, which can produce full-width letters, digits and spaces that look like ASCII on screen and are different characters. It is built from every text cell of the eight primary export sets, 3,504,720 cells in 359,974 rows, with the QA row Test123999 excluded. We hold no customer BOMs and describe no customer’s practice; the test below is one you can run yourself. Every part number is real.

Why does a full-width part number never match a Walsin export row?

Because the characters are different code points. We converted all 359,135 distinct part numbers to full-width letters, digits and punctuation, with the ideographic space U+3000 in place of the ASCII space, and 0 of them matched an export row as typed. The export side, read from 00_ALL_MLCC_parts_combined.xlsx and the other primary files, is clean: 0 of 359,974 part-number cells contain a character above U+007F.

We have not seen this counted anywhere. Part numbers are the one column that is perfectly ASCII, and they are also the column most likely to be re-typed. The 39,162 distinct part numbers with an embedded space are all chip resistor parts in the 64 non-empty chip resistor files, 39,087 with one space and 75 with two; our post on why Walsin part numbers do not paste into your BOM covers which space belongs to the part.

Which non-ASCII characters does the Walsin export actually contain?

Fourteen code points in the cells, and a fifteenth in the headers. 1,061,402 of 3,504,720 text cells (30.3 percent) hold at least one non-ASCII character, but three characters dominate: the plus-minus sign on 715,461 cells, Greek capital omega on 335,581 and the less-than-over-equal sign on 317,877.

The RF export’s column headers write the ohm in the headers with U+2126, the ohm sign, on 13 header cells, while every ohm in a data cell is U+03A9. A script that matches headers against cells by string will treat them as different. The inductor header uses U+03A9.

What key makes every Walsin part number unique after normalisation?

NFKC, then remove all whitespace, then upper-case. We built that key for all 359,135 distinct part numbers: it yields 359,135 distinct keys, so 0 collisions, and every adversarial variant (full-width, ideographic spaces, lower-case) resolved back to exactly one original.

The recipe, in one line of Python: re.sub(r"\s", "", unicodedata.normalize("NFKC", s)).upper(). NFKC folds full-width letters, digits and the full-width hyphen to ASCII, and folds the ideographic space to an ordinary space, which the next step removes. We checked that removing whitespace is safe in this catalogue: stripping the space from the 39,162 spaced part numbers collides with none of the other 319,973. Four part numbers hold a character that is not a letter, digit, hyphen or space: one inductor part that starts with an asterisk and three WQCW inductor parts with a vertical bar in Walsin_Inductor_ALL.xlsx. The key preserves them.

Keep the original string for ordering. The key is for finding the part, not for quoting it: the space in WF04A000 PTL is part of how the part is written in the export. A key that matches is a start, and the match still has to be checked field by field. That is what our free cross-reference tool does when you paste up to 500 part numbers or drop an Excel or CSV file.

Which BOM characters does Unicode normalisation not fix?

Four, and they matter for resistors. NFKC leaves the less-than-over-equal sign, the wave dash and a micro written as a letter u untouched, and it changes only 10,171 of 3,504,720 export cells, 10,162 of them the TCR tilde.

1. U+2264 versus U+2266. A BOM that says ≤ 100 with U+2264 does not text-match the 317,877 export cells that hold ≦±100 with U+2266. NFKC does not fold one into the other.
2. U+301C versus U+FF5E. The wave dash and the fullwidth tilde are separate characters. The export holds only the second, in 10,162 cells; NFKC turns it into an ASCII tilde, and 544 other TCR cells already hold 0 ~ 600 with spaces. The three TCR notations are counted in our post on chip resistor tolerance and TCR.
3. Micro written as u. 00_ALL_MLCC_parts_combined.xlsx writes microfarad as ASCII uF on 335 capacitance cells and the Unit column of Walsin_Inductor_ALL.xlsx writes uH on 644, so a BOM with U+00B5 or U+03BC will fold to the Greek mu and still miss both.
4. Applying NFKC to the export. It rewrites the 5 cells that hold U+02DA into a space plus a combining mark. Normalise your BOM, not the export.

What should a Japan BOM import check before it trusts a match?

Four steps, in order. Run the key, count the misses, split the misses by cause, and only then look at specifications. Across the 359,135 distinct part numbers we tested, the key misses nothing that exists in the export, so any remaining miss is a part or a spelling, not a character.

1. Key every row. Apply the NFKC, whitespace and case recipe above to your part number column. Rows that match now were character problems.
2. Count what is left. The remaining misses are either parts that are not in the export or parts with a different base number. Our post on crossing a 500-line BOM shows a miss list for a stock snapshot.
3. Normalise units, not just part numbers. Fold the ohm sign and the micro sign, and accept ≦ and ≤ as one.
4. Compare specification fields one at a time. A matching key proves the part exists, not that it fits.

We hold 28 Walsin datasheets, 451 pages, and the datasheet is the document to qualify against; ask for the sheet covering your part number and it will be in your inbox. For thin film, inductor, varistor, disc, RF and antenna we hold none, so confirm those lines on Walsin’s own site.

What is in Hong Kong is on ready stock: 1,125 part numbers as at 23 September 2026, a dated snapshot and not a live feed. The resistor range sits on the chip resistor page.

The short version

Every Walsin part number is ASCII, upper-case and unique: 0 of 359,974 cells hold a character above U+007F, and the NFKC, whitespace and upper-case key gives 359,135 keys for 359,135 part numbers. 1,061,402 of 3,504,720 text cells hold a non-ASCII character, and three characters dominate. NFKC fixes full-width input and the ohm sign but not ≤ against ≦, nor the wave dash, and it changes only 10,171 export cells. We will not tell you a part matches because its key does. We will show you the fields that differ. Qualification stays with you.

Send us the part number column of your BOM exactly as it leaves your spreadsheet and we will tell you how many rows are character problems and how many are real misses. And a question back: when your last import returned no match, did you check the part, or the characters?

Frequently asked questions